Overview of the Fourth-generation Mobile Communication System

Transcription

1 : Fourth-generation Wireless Access Overview of the Fourth-generation Mobile Communication System Narumi Umeda, Toru Otsu, and Tatsuro Masamura Abstract NTT DoCoMo is conducting research on the fourth-generation (4G) mobile communication system. This article outlines the technical issues involved in establishing such a system taking into consideration the capability and performance expected from future mobile communication systems. It also overviews the system configuration and discusses activities related to the standardization of 4G mobile communication systems. 1. Introduction The third-generation (3G) of mobile communication services, standardized as IMT-2000 (international mobile telecommunications 2000) [1], which began in October 2001 in Japan, have generated much anticipation regarding the development of a variety of multimedia services such as video communications. We believe that this will lead to mobile communications taking a more central role in our daily lives and to the expansion of this role as our lifestyles use mobile communications as a stepping stone to improve the quality of life over the next ten years. Such an era will require a fourth-generation (4G) mobile communication system that far surpasses the capability of IMT Providing mobile communication services based on new technology involves more than simply proposing and proving technology it also requires field-testing of functions and performance, standardization of technical specifications, development of mobile terminals, and construction of network facilities. New mobile communication services will require more time and effort to establish than other types of communication services do. Research and development for the timely introduction of a 4G system that has the performance required to serve as a part of the future foundation for our society and lifestyle is in progress in NTT DoCoMo. NTT DoCoMo, Inc. Yokosuka-shi, Japan umedan@nttdocomo.co.jp In this article, we describe a basic approach to the technical issues and system configuration involved in achieving the capabilities and performance required of the 4G system based on the research at NTT DoCoMo. We also describe the trends in standardization concerning future mobile communication systems. 2. System objectives 2.1 Requirements (1) Broadband communications Up until now, the traffic carried by mobile communication systems has mainly been voice communications. The second-generation (2G) system, the personal digital cellular (PDC) system, introduced i- mode services [2] that have brought about the currently popular form of Internet access, electronic commerce, and , which are mainly text-based data communications via a cellular phone. The IMT system offers high bit-rate transmission services from 64 to 384 kbit/s, and the proportion of data to voice traffic is expected to increase. Moreover, the rising popularity of broadband services such as ADSL (asymmetric digital subscriber line) and optical fiber access systems and office and home LANs is likely to lead to a demand for comparable services in the mobile communications environment. (2) Low cost To make broadband services available so that users can exchange various kinds of information, it is necessary to lower charges dramatically to keep the cost 12 NTT Technical Review

2 Technologies at or below that of existing services. The IMT-2000 system aimed at lowering the bit cost and establishing economical rates, but the 4G system requires a broadband channel and an even lower bit cost. (3) Wide service area One feature of mobile communications is that it is available for use anytime and anywhere. These capabilities are also important for future mobile communications. When a new system is first introduced, it is generally difficult to provide such an extensive service area as the existing system, but customers will not buy the new terminals if they have restricted service areas. Moreover, to support terminals that have relatively large display screens, such as personal digital assistants (PDAs) and personal computers with wireless capability, especially ones used with advanced services, which will often be used indoors, we need to provide better coverage of indoor service areas. (4) Diversified services and ease of use The target subscriber base for mobile communications comprises various types of users. In the future, we expect to enhance the system performance and functionality to introduce a variety of services that include not only ordinary telephone services, but also services that transfer information utilizing all five senses. These services must be made easier for anyone to use. 2.2 Design objectives The design objectives for meeting the above requirements are shown in Fig. 1. Considering that video and data communications will be the main features, the 4G system must provide even higher transmission rates and larger capacity (i.e., both number of users and traffic volume) than IMT Also, considering that the video transmission quality in current broadcasting is achieved by a transmission rate of several megabits per second, that LAN transmission rates are from 10 to 100 Mbit/s, and that the rate of ADSL is several megabits per second, the design objective is a transfer rate of approximately 100 Mbit/s in an outdoor mobile environment and gigabitclass rates indoors. It will not be possible to accommodate future mobile communication traffic unless a transmission capacity of at least ten times that of the IMT-2000 is achieved. To ensure throughput for communications between terminals and achieve highlevel realtime communications, it is necessary to achieve a low transfer delay time of 50 ms. Also, assuming that future services will be based on Internet protocol (IP) networks, efficient transmission of IP packets over wireless connections is also a necessity. While increased capacity is also effective in lowering the bit cost, the cost per bit must be reduced to between 1/10 and 1/100 of the current levels by reducing the infrastructure equipment, operation, and construction costs. The design objectives described above focus on services that have higher performance than existing services, yet are easy to use. It is necessary to pioneer new markets by making use of the capabilities and performance of the 4G system, such as integration with indoor wireless LAN and wired systems, and by implementing a mechanism for introducing new services in a short time. 3. Basic approach to 4G system configuration 3.1 Technical issues The technical issues concerning wireless technolo- Capability and performance Start new services based on new capabilities Enhance services through performance enhancements Information bit rate: 384 kbit/s System capacity Cost Base station network transport system: ATM New service platform -Rapid deployment of new services -Easy development of new services Seamless connection and handoff between heterogeneous access systems 100 Mbit/s (peak rate in mobile environment) 1 Gbit/s (peak rate in indoor environment) 10 times that of 3G 1/10 to 1/100 per bit All-IP Transmission delay time 50 ms or less IMT G system (2010) Fig. 1. Design objectives. Vol. 2 No. 9 Sep

3 gy that need to be addressed to achieve the system objectives described above are shown in Fig. 2. (1) High-capacity and high-rate transmission IMT-2000, which employs wideband code division multiple access (W-CDMA), achieves a transmission rate of 2 Mbit/s with a 5-MHz frequency bandwidth. Furthermore, technology for transmission at approximately 10 Mbit/s with the same frequency bandwidth using multi-level adaptive modulation and demodulation is under development [3]. To achieve rates of 100 Mbit/s to 1 Gbit/s, we must use a larger frequency bandwidth and new transmission systems that are suited to high-rate transmission. For data communications, we will need a radio access system that can transmit packets efficiently. Considering the importance of indoor area coverage in the future, technologies that allow use both indoors and outdoors must also be developed. To obtain the broadband frequencies for achieving high-rate transmission and meet the expected large increase in data traffic demand, we must consider new frequency bands and we must develop the circuit technology needed to make amplifiers and filters and understand radio wave propagation in these bands. At the same time, technology for making efficient use of limited spectrum resources is also important. (2) Lower costs With conventional system configuration technology, using a higher frequency band to achieve a higher transmission rate generally reduces the area of the cell that one base station can cover. Retaining the original coverage area requires more base stations and increases the network cost. To avoid that problem, it is necessary to expand the cell radius by using higher performance radio transmission and circuit technology such as improved modulation/demodulation techniques that can cope with a low signal-tonoise ratio, adaptive array antennas, and low-noise receivers. To further reduce the system construction and operating costs, we must study diversified entrance links that connect base stations to the backbone network, autonomous base station control technology, and multi-hop radio connection technology employing simple relay stations. (3) Interconnection based on IP networking One way to ensure that users of the new system do not have restricted service areas is to ensure that new terminals can handle the existing system as well as the new one. Moreover, considering the demand for international roaming, a terminal that can be configured to work with multiple systems based on software defined radio (SDR) technology [4] is an effective way to cope with introductory periods and differences in operating frequency bands among different countries and regions. Furthermore, future mobile communication networks will be integrated with heterogeneous access methods and various kinds of cells with interconnection capabilities based on IP net- Signal transmission and antenna technology High-rate transmission/high-performance modulation and demodulation High-capacity, broadband packet access Adaptive array antenna High-performance FEC, ARQ, TPC, diversity reception Radio wave propagation and radio link budget design Microwave mobile propagation Broadband propagation Indoor propagation RAN architecture and control IP packet control/distribution control RAN QoS packet transmission control Microcell configuration Seamless control for connection to other networks Entrance link RF circuit/emc Highly efficient amplifier Super-low-noise receiver Software defined radio Microwave EMC Optical link Entrance links Broadband wireless links Integrated optical/wireless network Microcell FEC: forward error correction ARQ: automatic retransmission request TPC: transmission power control RAN: radio access network QoS: quality of service RF: radio frequency EMC: electromagnetic compatibility Fig. 2. Technical issues concerning wireless technology. 14 NTT Technical Review

4 working. Accordingly, interconnection and handover between such various access systems are required in addition to handover and roaming within one mobile communication system. 3.2 System configuration (1) IP-based connection configuration The 4G system will be configured for connection to IP networks, considering efficient transmission of IP packets, co-existence with other access systems, ease of system introduction, expandability, and other such factors. IP networks can also connect with or accommodate wireless access systems other than 4G systems. The 4G wireless access point (hereinafter 4G- AP) will be connected to an access router (AR) as shown in Fig. 3 and will have wireless control functions for wireless transmission, handover, etc., allowing communication with mobile nodes operating on IP. The 4G-APs will form their respective cells. When a mobile node moves between cells, handover will be accomplished by simply switching access points and wireless areas if the two 4G-APs are connected to the same AR. If the 4G-APs belong to different ARs, then the packet transmission route on the IP network must be changed rapidly. The cooperative operation of 4G-AP switching and IP routing is important for smooth handover. For handover between a 4G-AP and an AP of another system, the mobile node must have functions for accessing both systems. Handover will be performed by monitoring and comparing different systems to select the one that is more suitable. (2) Cell classification and configuration according to communication environment The 4G system has cells for outdoors, indoors, and inside moving vehicles, as shown in Fig. 3. Outdoor cells cover a wide area, unlike the hotspot areas of wireless LANs, and allow high-rate packet transfer for fast-moving terminals. Indoor areas are covered by indoor APs, because the radio waves to/from outdoor base stations suffer large attenuation. Indoor APs are designed not only to provide a high rate transfer and simple operation, but also to compete with expected future wireless LANs. Furthermore, cells within moving vehicles such as buses and trains (moving cells/networks) are served by a mobile router (MR) that has wireless functions and relays signals between a base station and each terminal in the vehicle, rather than the terminals individually communicating directly with the base station in the conventional method. This configuration is designed to achieve efficiency in terms of terminal transmission power, transmission rate, control signal volume, etc. A multi-hop connection, which is effective in expanding the cell size, is being investigated as a way to overcome dead spots caused by shadowing. Data transmission via relay stations is expected to allow communications even when the effects of limited terminal transmitting power and radio wave propagation attenuation are large, as shown in Fig. 4 [5]. (3) Multimedia communications Conventional IP networks have provided mainly best-effort services, but with realtime applications expected to increase as multimedia communication To other networks (Internet, etc.) 4G-AP IP network AR Other access Indoor cell Outdoor cell 4G-AP 4G-AP 4G-AP 4G-AP Other access 4G-AP MR Moving cell AR: access router 4G-AP: 4G wireless access point MR: mobile router Fig. 3. 4G system configuration. Vol. 2 No. 9 Sep

5 Direct communication is blocked by the building. Direct communication is not possible because the distance exceeds the single-hop range. Building Base station Multi-hop Single-hop Multi-hop Fig. 4. Communication problems overcome by multi-hop connection. diversifies, the importance of services that take into account quality of service (QoS) is also expected to increase. The 4G system configuration allows for a mechanism that guarantees the transmission rate to some extent and that prioritizes packet transfer by the packet type in cooperation with the IP network for QoS-aware packet transmission on a mobile radio link, which is the bottleneck. 4. Trends in standardization 4.1 ITU-R activities In 2000, the year in which the prospect of introducing the IMT-2000 system came into view, the international telecommunication union (ITU) began research on future development of IMT-2000 and other systems. In the ITU radiocommunication sector (ITU-R), investigation of Q.229/8 on future development of IMT-2000 and systems beyond IMT-2000 was assigned to study group 8 (SG8) working party 8F (WP8F), which was established in November 1999, and work on this topic began in March At the world radiocommunication conference held in June 2000 (WRC-2000), ITU-R resolved to conduct research on future systems, including spectrum requirements, to investigate the research situation at WRC-2003, and to review spectrum requirements at subsequent WRCs. ITU-R WP8F formulated a recommendation regarding a future vision to give direction to future technological development. The recommendation was approved at the February 2003 meeting of SG8 and forwarded to a higher-level organization, the radiocommunication assembly (RA). In RA, it was approved as the framework recommendation in June In WRC-2003 held in July, approval was given for the agenda items of WRC-2007 to include the frequency assignment for systems beyond IMT In that recommendation, systems beyond IMT-2000 is considered to cover all future mobile communication systems, including the current IMT and its enhanced versions. The various wireless access systems will need to cooperate via the network so that users can use the full range of capabilities of the systems beyond IMT-2000 without being aware of individual wireless access systems. Furthermore, there is now recognition of the need for a new wireless access system and a frequency band for it to operate in to cover the performance region that cannot be achieved by advanced IMT-2000 systems (transmission rates of approximately 100 Mbit/s during highspeed movement and approximately 1 Gbit/s when not moving, although these bit rates assume sharing by users and the specific values are research targets). Furthermore, the target time for implementation of the new wireless access system is 2010 [6]. In the future, the study of spectrum requirements and research on specific technological issues are expected to make progress. 4.2 Activities in Japan In Japan, the New Generation Mobile Communication Committee was established in the Information and Telecom Council of the Ministry of Public Management, Home Affairs, Posts and Telecommunications between October 2000 and June 2001, the same period during which ITU began research, and it formulated a vision for future mobile communications [7]. The results were reflected in ITU-R s vision recommendation described above. Furthermore, based on the committee s findings, the mobile IT Forum (mitf) was established in June 2001 to investigate and research the early implementation of the 4G system, mobile commerce, and other topics. [8]. 4.3 Activities in other organizations The Wireless World Research Forum (WWRF), an 16 NTT Technical Review

6 organization of mainly European vendors, is also conducting research concerning a future vision for wireless communications [9], and new research projects are being organized on the basis of the results. In Europe, the 6th Framework Programme is being promoted. It established Ambient Network [10] and WINNER [11] this January and initiated research on future systems. 4.4 Future plans The vision for systems beyond IMT-2000 proposed so far by ITU-R coincides with the research objectives of NTT DoCoMo. We will continue to contribute to the standardization work being accomplished by ITU-R in coordination with future progress in research. It is also important to make global standards through cooperation with other research organizations. IEEE, which has established wireless LAN standards, has begun to study nextgeneration wireless LANs. Considering the expected increase in indoor use of the 4G system and the growing affinity for the Internet, further cooperation with IEEE standardization is necessary. 5. Conclusion We outlined research projects toward the 4G system. We described system requirements, topics for study, and a basic approach to the system configuration. We also presented trends related to standardization in this field. References [1] [2] [3] F. Adachi and M. Uesugi, Latest Trends in CDMA Technology, IEICE Journal, Vol. 86, No. 2, pp , [4] K. Uehara, K. Araki, and M. Umehira, Trends in Research and Development of Software Defined Radio, NTT Technical Review, Vol. 1, No. 4. pp , [5] A. Fujiwara, S. Takeda, H. Yoshino, T. Otsu, and Y. Yamao, System Capacity Expansion Effect of Multi-hop Access in a Broadband CDMA Cellular System, IEICE Trans. B, Vol. J85-B, No. 12, pp , [6] T. Otsu, The Challenge of Systems beyond IMT-2000 approach from wireless, ITU Journal, Vol. 33, No. 3, pp , [7] bunkakai/abstract.pdf [8] [9] [10] [11] Narumi Umeda Director, Communication Systems Laboratory, Wireless Laboratories, NTT DoCoMo, Inc. He received the B.E. and M.E. degrees in electronic engineering from Hokkaido University, Sapporo, Hokkaido in 1985 and 1987, respectively. He joined NTT Laboratories in In 1992, he transferred to NTT Mobile Communications Network, Inc. (now NTT DoCoMo, Inc.). He has been engaged in research on the radio link control for the personal digital cellular (PDC) system and IMT His current research interests are in the radio control for 4G mobile communications systems (beyond 3G). He was a co-recipient of the Japan Institute of Invention and Innovation (JIII) Imperial Invention Prize in 1998 and the best paper award of the International Conference on Telecommunications at ICT2002. He is a member of the Institute of Electronics, Information and Communication Engineers (IEICE) and IEEE. Toru Otsu Director, Global Business Department, NTT DoCoMo, Inc. (Formerly with the Wireless Laboratories) He received the B.E., M.E., and Ph.D. degrees in global information and telecommunications from Waseda University, Tokyo in 1983, 1985, and 2002, respectively. In 1985, he joined NTT and in 2000 he transferred to NTT DoCoMo, Inc. From 1992 to 1993, he was a visiting researcher at ENST (Ecole Nationale Superieure des Telecommunications), France. Since joining NTT DoCoMo, he has been involved in research projects on 4G mobile communications systems and related standardization activities. His research interests are network architectures, network control protocols, and resource management in mobile and satellite networks. He is a member of IEICE and IEEE. Tatsuro Masamura Associate Senior Vice President, Managing Director of Wireless Laboratories, NTT DoCo- Mo, Inc. He received the B.E. and M.E. degrees in electrical engineering from the University of Tokyo, Tokyo in 1974 and 1976, respectively. He joined the Electrical Communication Laboratories, Nippon Telegraph and Telephone Public Corporation (now NTT) in Since then he has been engaged in R&D of satellite communication systems based on CDMA and TDMA. He worked at the Communication Research Center, Department of Communication of Canada as an exchange scientist from 1983 to From 1992 to 1994, he was a manager of satellite communications systems at NTT Network Development Center. From 1994 to 1999, he was engaged in the development of multimedia services using satellite communications. From 1999 to 2002, he was the executive manager of NTT Network Innovation Laboratories. Since April 2002, he has been with NTT DoCoMo as the Managing Director of Wireless Laboratories responsible for R&D activities for next-generation mobile communications. Vol. 2 No. 9 Sep

7 Broadband Packet Wireless Access Mamoru Sawahashi, Sadayuki Abeta, Hiroyuki Atarashi, Kenichi Higuchi, Motohiro Tanno, and Taisuke Ihara Abstract Packet access in the radio access network (RAN) is essential to reduce network costs and achieve highcapacity multimedia mobile communications capable of supporting services such as high-resolution video communication. This article presents targets and key techniques for the broadband packet wireless access system that we propose, which enables seamless support of both cellular systems and isolatedcell environments, such as hotspot areas and indoor offices, using the same air interface for efficient packet access. It also describes efficient packet access techniques for guaranteeing the quality of service (QoS), including handover in the RAN on both the data-link layer and the physical layer. 1. Introduction A maximum information bit rate of 2 Mbit/s or greater has been specified as the requirement for third-generation (3G) mobile communications called international mobile telecommunications-2000 (IMT-2000) in the radiocommunication sector of the international telecommunication union (ITU-R). We have experimentally demonstrated high-speed data transmission above 2 Mbit/s with high quality at an average bit error rate of less than 10 6 using a 5-MHz frequency bandwidth in the downlink, applying three-code multiplexing with a spreading factor (SF) of 4 using two-branch antenna diversity reception. Meanwhile, the third-generation partnership project (3GPP) has completed specifications focusing on high-speed packet data transmission in the downlink called high-speed downlink packet access (HSDPA) based on an air interface using wideband code division multiple access (W-CDMA) [1]. In HSDPA, a much higher peak throughput than 2 Mbit/s is possible with low latency and high efficiency by employing adaptive modulation and channel coding (AMC) including 16 quadrature amplitude modulation NTT DoCoMo, Inc. Yokosuka-shi, Japan sawahashi@nttdocomo.co.jp (QAM), hybrid automatic repeat request (H-ARQ) with packet-combining in the medium access control (MAC) layer, and fast packet scheduling (i.e., user diversity). Moreover, we have shown that a peak throughput of nearly 10 Mbit/s is possible with a 5- MHz bandwidth by using a multipath interference canceller or chip equalizer even in a multipath fading channel. However, to meet the current huge increases in the amount of data traffic and further ones expected in the future, new broadband wireless access methods must provide broadband packet transmission with a maximum throughput above 100 Mbit/s in the downlink using a bandwidth of approximately 50 to 100 MHz (note that the target throughput corresponds to an approximately 10-fold increase over that achievable in HSDPA with a 5-MHz bandwidth). In addition, this broadband wireless access must flexibly support both isolated-cell environments (such as hotspot areas and indoor offices) and cellular systems from the standpoint of further reducing the cost of radio access networks (RANs). Furthermore, since it is presumed that the signal format in the wireless channel is a packet format, the service being offered is basically a best-effort type according to the channel condition of each user and traffic conditions within the cell, where minimum throughput is guaranteed with the required quality of service (QoS), e.g., delay and residual packet error rate. 18 NTT Technical Review

8 Cellular system Hotspot area Supported by the same air interface IP-based core network (IP: Internet protocol) Fig. 1. Concept of the broadband packet wireless access system supporting a cellular system and local areas by the same air interface. In this article, we present targets for a broadband packet wireless access system, which is for the systems beyond IMT-2000, and overview key techniques for it. Section 2 describes the targets of broadband packet wireless access. Section 3 then explains the RAN configuration that we propose, and section 4 presents broadband packet wireless access techniques focusing on wireless access schemes. After that, section 5 briefly presents efficient packet access techniques, and section 6 discusses multiple-antenna transmission methods, which are essential for enabling wider coverage area and improving the achievable throughput. 2. Targets of broadband wireless access Figure 1 shows the concept of the proposed broadband packet wireless access system. In the future RAN, further decrease in the network cost is a very important requirement for offering rich multimedia services to customers via wireless communications. Therefore, the proposed concept can support both a cellular system with a multi-cell configuration and local areas (such as hotspot areas and indoor office environments) by the same air interface (i.e., the same carrier frequency, frequency bandwidth, radio frame structures, etc.). Only the main radio parameters such as spreading factor, data modulation scheme, and channel coding rate need to be changed. By changing them according to the cell configuration, channel load, and channel condition of each user, we can achieve the maximum system capacity based on the same air interface, that is, the same broadband wireless access scheme. In our broadband packet wireless access system, the target of the peak throughput for the downlink in a cellular environment is 100 Mbit/s, which is approximately ten times the peak throughput forecast for HSDPA using a 5-MHz frequency bandwidth. Achieving a throughput of 100 Mbit/s will of course help to reduce network cost, but it will also enable largecapacity data downloads and high-resolution video communication services among multiple users. However, local areas (generally, short-distance and shorttime-dispersion environments) will probably require throughput in excess of 100 Mbit/s to support large amounts of data traffic. We therefore aim for a maximum throughput of 1 Gbit/s (frequency efficiency: 10 bit/s/hz) for these types of areas and intend to support throughput flexibly for various environments using the same air interface. For the uplink, the target for frequency efficiency for hotspots and indoor environments is 7 bit/s/hz. This comes to a maximum throughput of approximately 300 Mbit/s for a frequency bandwidth of 40 MHz. For a cellular environment, the target is a maximum throughput of 20 Mbit/s per sector *1. In the broadband wireless access system, all signals are transmitted by a packet configuration in the RAN, i.e., to transmit both realtime and non-realtime traffic data as packet signals. Here, channels can be allocated to multiple users in a cell on the basis of time-division multiplexing using a shared-channel packet format. This will significantly reduce the number of physical radio devices making up base-station equipment compared with the circuit-switching transmission system that allocates a dedicated physical channel to each user. *1 Sector: Cells are split into sectors for greater efficiency. Vol. 2 No. 9 Sep

9 3. Radio access network configuration in broadband wireless access 3.1 Inter-cell synchronization A broadband wireless access system must be able to support a seamless service area covering not only a cellular environment but also hotspots in underground shopping centers, airports, hotel lobbies, and indoor environments like offices. To achieve flexible deployment, especially to indoor environments, an inter-cell asynchronous system is advantageous over an inter-cell synchronous system based on time synchronization. A mobile terminal needs to establish a wireless link with the base station that provides the minimum path loss. In a wireless access system that supports a multi-cell cellular system, this criterion is nearly the same as the criterion of finding the reference signal in the downlink with the highest received power level (in W-CDMA, the common pilot channel is used for this purpose). Thus, in W-CDMA, the cell or sector whose common pilot channel in the downlink has the highest received power is selected as the optimal cell/sector. The situation is somewhat different, however, in a broadband wireless access system that supports both cellular and isolated cells, as shown in Fig. 2. Specifically, the base-station transmission power of cellular cells differs from that of hotspot cells, which means that the cell with the highest received power level of the common pilot channel in the downlink is not necessarily the cell with the lowest path loss between the base station and mobile terminal. In Received power of common pilot channel P cellular, P hotspot Cellular cell (outdoors) Path loss (L cellular ) P cellular > P hotspot L cellular < L hotspot Hotspot cell (indoors) P cellular < P hotspot L cellular > L hotspot Path loss (L hotspot ) P cellular : cellular cell received power P hotspot : hotspot cell received power Fig. 2. Cell selection when cellular cells and hotspot cells coexist. the example shown in the figure, the received power of the pilot channel from an outdoor cellular cell is highest, but the cell with the lowest path loss is an indoor hotspot. Establishing a wireless link with the indoor cell leads to the maximum reduction in the transmission power required by the mobile terminal. In the manner described above, a search must be performed for the cell/sector having the lowest path loss between the base station and mobile terminal in a system where isolated environments coexist with a cellular environment. This cell search can be divided into three types: (1) initial cell search, (2) cell search in active mode, and (3) cell search in idle mode. For cell searches in the active and idle modes, and standby searches, the currently connected cell (sector) informs the mobile terminal about the transmission power and cell type (cellular or isolated) of neighboring cells so that the terminal can find the optimal cell having the lowest path loss between itself and the base station. For the initial cell search, two methods can be considered. In the first method, which is similar to the method used in conventional cellular systems, the mobile terminal first establishes a wireless link with the cell whose downlink pilot channel has the highest received power. The terminal then receives and decodes a report on neighboring cells from that cell. If there is a neighboring cell with lower path loss than the currently connected cell, the terminal switches its wireless link to it. In the second method, each cell is allocated beforehand a unique scrambling code indicating a cellular or hotspot cell [2]. As a result, a mobile terminal in an initial cell-search stage can identify the cell type and thus search for the optimal cell with the lowest path loss simply using the processing in the physical layer. Of these three types of cell searches, the initial cell search takes the longest time. To reduce the search time, three-step cell search algorithms have been proposed using orthogonal frequency and code division multiplexing (OFCDM) or orthogonal frequency division multiplexing (OFDM) [3], [4]. In the first step, the system detects the OFCDM or OFDM symbol timing. In the second step, it detects the group to which a cell-specific scrambling code belongs as well as the frame/slot timing. To detect this information, a method that uses a synchronization channel [3] and one that uses a time-multiplexed common pilot channel [4] have been proposed. In the third and final step, the system detects the scrambling code of the optimal cell among the scrambling-code candidates belonging to the scrambling-code group detected in the sec- 20 NTT Technical Review

10 IP-based core network Router Cell 1 Cell 2 Inter-sector handover soft handover (fast sector selection) Inter-cell handover hard handover Fig. 3. Handover control. ond step. 3.2 Handover In addition to handover between cells/sectors in a cellular system, a function for achieving handover between a cellular cell and an isolated cell is essential to achieving seamless cell deployment. A handover operation on the data-link or lower layer is called inter-cell or inter-sector macro-diversity *2, and the required QoS on these layers is guaranteed by employing the macro-diversity. On higher application layers, both the data delay via wired transmission based on the Internet protocol (IP) network in the system and the control signaling delay corresponding to the mobility of the mobile terminal must be taken into consideration. In W-CDMA, the mobile terminal sets up dedicated channels with multiple cells/sectors and obtains a diversity effect by performing a soft handover in which the same information is transmitted to the mobile terminal via multiple cells/sectors on the downlink and received by multiple cells/sectors on the uplink. This macro-diversity effect between cells/sectors in a soft handover results in high-quality transmission in circuit-switching mode. A broadband wireless access system, however, employs transmission-slot allocation (packet scheduling) and ARQ, which means that an optimal handover algorithm on the data-link layer or below must take into account the effects of packet scheduling and ARQ. In this regard, a study has been performed on *2 Macro-diversity: Improvement obtained by combining the signals received at multiple cells (or sectors). throughput performance versus cell selection interval during inter-cell macro-diversity reception for OFCDM on the downlink [5]. For a cell selection interval of more than 500 ms, cell-selection updating cannot follow the shadowing variation. This results in the selection of a cell with a small signal-to-interference power ratio (SIR), i.e., a large path loss, and an increase in the required ratio of the signal energy per bit to noise power spectrum density (E b /N 0 ). Figure 3 shows the proposed method of handover control for the data-link layer and below. When taking into account fast packet scheduling and ARQ, fast hard handover with a cell selection interval of about 100 ms is considered to be appropriate for inter-cell handover. A hard handover can simplify the control of the core network between different cells compared with a soft handover. As for inter-sector handover, processing within the base station makes it easier to distribute and combine data sequences. A soft handover (including fast sector selection on the downlink) is therefore expected to have a traffic-dispersion effect when traffic becomes concentrated in a particular sector. 4. Broadband wireless access schemes 4.1 Duplexing Two well-known duplexing schemes are frequency division duplexing (FDD) and time division duplexing (TDD). TDD has the advantage of not requiring a pair of frequency bands, although it does require base stations to be synchronized (frame synchronization) for application to a cellular system. In an isolated-cell Vol. 2 No. 9 Sep

11 f 1 Frequency f A f B f C Frequency System band System band (a) One-cell frequency reuse (b) Three-cell frequency reuse Fig. 4. Comparison of one-cell and three-cell frequency reuse. environment, moreover, TDD can perform flexible slot allocation for each link by varying the transmission/reception slot-allocation ratio on the uplink and downlink using the same frequency band. However, in a cellular system, uplink/downlink slot allocation ratio needs to be same among a cluster of neighboring cells. FDD, on the other hand, while requiring a pair of frequency bands for the uplink and downlink, does not require base-station synchronization, so it is considered to be preferable to TDD from the viewpoint of flexible cell deployment. While both will probably have to be supported in the end, a system configuration based on a single wireless interface (frame configuration) is desirable. 4.2 Cell frequency reuse Figure 4 shows the frequency usage efficiency in a cellular system for one-cell and three-cell frequency reuse. With one-cell frequency reuse (Fig. 4(a)), the transmitted signals are spread over the whole system band by spreading in the time or frequency domain while the despreading process decreases interference and noise components by the inverse of the spreading factor on average, which means that a spreading gain can be expected. In ordinary time division multiple access and frequency division multiple access, however, three-cell frequency reuse (Fig. 4(b)) is necessary to reduce co-channel interference. As a result, the frequency bandwidth per cell/sector is 1/3 of the full system band. In short, one-cell frequency reuse is an essential condition for increasing system capacity in a cellular system. Cumulative distribution (%) One-cell frequency reuse Space attenuation: 3.76 power law Shadowing: lognormal distribution (standard deviation: 8 db) Average received SIR Three-cell frequency reuse 1 sector 3 sectors Fig. 5. Cumulative distributions of average received SIR for one-cell and three-cell frequency reuse. Figure 5 shows cumulative-distribution of average received SIR (equal to the ratio of desired-signal power to interference power from neighboring cells in this case) for one-cell and three-cell frequency reuse on the downlink of a cellular system. It is assumed here that the cell radius is the same for all cells and that the transmission power is the same for all sectors. Distance-dependent path loss follows a 3.76 power law, while shadowing variation follows a lognormal distribution with a standard deviation of 8 22 NTT Technical Review

12 db. The 50% values of these cumulative-distribution plots show that the received SIR for one-cell frequency reuse is about 10 db less than that for three-cell frequency reuse, which means that one-cell frequency reuse needs gain to suppress about 10 db of interference from neighboring cells. Spreading of the time or frequency domain can easily achieve one-cell frequency reuse by effectively reducing interference from neighboring cells. In addition, to prevent each symbol from being affected by same-pattern interference from neighboring cells in one-cell frequency reuse, it is essential to employ cell-specific (or userspecific) scrambling code. 4.3 Downlink wireless access scheme: VSF-OFCDM Table 1 compares broadband wireless access systems on the downlink and Fig. 6 compares DS- CDMA and multicarrier transmission systems on a broadband channel. These systems can be roughly divided into those that perform spreading using cellspecific spreading codes and those that do not. In a broadband wireless channel, a large number of delayed paths are observed. If spreading is applied to the broadband wireless acess scheme based on the time-domain-spreading approach taken by DS- CDMA, the number of paths that can be resolved increases. However, at the same time, paths having different delay times generate interference (multipath interference), which offsets the Rake diversity effect. Table 1. Comparison of wireless access systems on the downlink. Access system Single/multicarrier DS-CDMA OFCDM OFDM Single/multicarrier TDMA Spreading Yes Yes No No Number of carriers 1 or a few Many Many 1 or a few Effects of multipath interference Degrades the Rake diversity effect Robust against multipath interference Robust against multipath interference Degrades signal due to inter-symbol interference Frequency reuse One-cell frequency reuse One-cell frequency reuse Cell frequency reuse basically needed Cell frequency reuse basically needed DS-CDMA system Path #1 Multicarrier transmission system (OFCDM, OFDM) Path #1 Path #2 Path #2 Path #3 Path #3 Rake combining Multipath interference component Guard interval (>maximum delay time of multipath) Frequency Frequency Fig. 6. Comparison of wireless access systems in a broadband channel. Vol. 2 No. 9 Sep

13 If the SIR of each despread path is very small, the signal after Rake combining cannot achieve the required SIR. Therefore, a wireless access system robust against multipath interference is therefore essential to achieving high-quality signal transmission in a broadband channel. In this regard, a multicarrier transmission scheme first subjects a high-speed data stream to a serial-parallel conversion generating many subcarrier signals whose symbol length is sufficiently longer than the multipath propagation delay time. It then proceeds to transmit these low-symbolrate data streams in parallel. In other words, in multicarrier transmission, the frequency band is divided into a large number of narrow-band signals resulting in a small bandwidth per subcarrier. This means that amplitude and phase variations within a subcarrier can be treated as flat fading and that the effects of waveform distortion caused by frequency-selective fading can be reduced. Even if there is a subcarrier whose received power has dropped due to fading, its decoded error can be compensated for by applying error-correction channel coding across multiple subcarriers whose received power has not dropped, resulting in high-quality reception. On the downlink of a cellular system, OFCDM, which is based on multi-carrier CDMA (MC-CDMA) [6], [7], can reduce the effects of multipath interference while obtaining a frequency diversity effect by spreading and mapping channel-coded symbols across multiple subcarriers. For this reason, OFCDM is more suitable for increasing capacity than other wireless access systems [8], [9]. Thus, in addition to being robust against multipath fading, which is a characteristic of multicarriers, OFCDM can achieve one-cell frequency reuse in a flexible manner and increase system capacity by applying spreading in either the time or frequency domain. In an isolatedcell environment, however, interference from neighboring cells is small and system capacity can generally be increased by not using spreading. The reason for this is that when spreading in the frequency domain is used, code orthogonality among multiplexed code channels collapses due to frequencyselective fading caused by multipath interference. Similarly, when spreading in the time domain is used, amplitude fluctuation occurs in the time domain in a high-speed Doppler environment, giving rise to intercode interference that in turn destroys the orthogonality in the time domain. In either case, it is difficult to multiplex as many code channels as corresponds to the spreading factor. To deal with the above situation, OFCDM with a variable spreading factor (VSF-OFCDM) has been proposed [10], [11]. This system adapts the variable spreading factor in the OFCDM time and frequency domains according to the cell configuration, propagation conditions (delay spread, maximum Doppler frequency, and interference from other cells), channel load, radio parameters (data modulation scheme and channel coding rate), and other conditions. Figure 7 shows the concept of VSF-OFCDM with two-dimensional spreading. Here, one data modulation symbol Code Time domain spreading (SF Time ) Time Frequency domain spreading (SF Freq ) User 1 User 2 User 3 Total spreading factor SF > 1 Frequency Time-domain-spreading only (SF Freq = 1) Cellular system Control of variable spreading factors (SF Time, SF Freq ) Hotspot indoor cell Fig. 7. Concept of VSF-OFCDM using two-dimensional spreading. 24 NTT Technical Review

14 Table 2. Comparison of wireless access systems on the uplink. Access system Single/multicarrier DS-CDMA OFCDM OFDM Single/multicarrier TDMA Spreading Yes Yes No No Number of carriers Low power-consuming terminal 1 or a few Many Many 1 or a few Low peak-power transmission possible by spreading Large peak power Large peak power Large peak power Large capacity Performance improved by the Rake diversity Accuracy of channel estimation degrades Accuracy of channel estimation degrades Signal degrades due to inter-symbol interference Frequency reuse One-cell frequency reuse One-cell frequency reuse Cell frequency reuse basically needed Cell frequency reuse basically needed will be spread across several consecutive OFCDM symbols and several consecutive subcarriers (these numbers are given by SF Time and SF Freq, which denote the spreading factor in the time and frequency domains, respectively) with the total spreading factor represented as the product SF = SF Time SF Freq. As indicated in Fig. 7, two-dimensional VSF-OFCDM can (i) control the total spreading factor in accordance with the cell configuration (the mobile terminal sets the spreading factor based on control information from the base station), (ii) adaptively control SF Time and SF Freq in accordance with propagation conditions, channel load, and radio parameters, and (iii) maximize channel capacity for both a cellular system and an isolated-cell environment. 4.4 Uplink wireless access scheme: VSCRF-CDMA Table 2 compares broadband wireless access systems on the uplink. From the viewpoint of low mobile-terminal power consumption, the most important requirement on the uplink, the DS-CDMA approach is more advantageous than the multicarrier approach as in OFDM and OFCDM, which use many subcarriers having a large peak-to-average power ratio. The uplink, moreover, requires a dedicated pilot channel for each mobile-terminal physical channel to perform coherent detection and demodulation. Consequently, for OFDM and OFCDM, channel estimation must be performed for each subcarrier, and for the same pilot power condition, the pilot channel signal power per subcarrier is small compared with that of DS-CDMA. As a result, the DS-CDMA system has been reported to maintain more accurate channelestimation, to reduce the received E b /N 0 that satisfies the required packet error rate compared with multicarrier systems, and to be capable of increasing the link capacity [12]. Moreover, in DS-CDMA, there is a bandwidth that can minimize the required transmission power (i.e., received E b /N 0 ). This optimal subcarrier bandwidth is determined based on the tradeoff between the Rake diversity and increasing multipath interference. When propagation conditions (such as delay profile shape and number of paths) and the spreading factor are given as parameters, the received E b /N 0 required can be reduced the most by a subcarrier bandwidth from 20 to 40 MHz [12]. Accordingly, a multicarrier/ds-cdma system configured on the basis of this optimal subcarrier bandwidth in accordance with the system band is a promising wireless access system from the viewpoint of link capacity. The DS-CDMA system can achieve one-cell frequency reuse in a flexible manner through spreading in the time domain. The advantage of one-cell frequency reuse diminishes, however, in an isolated-cell environment for which interference from neighboring cells is small. Here, link capacity turns out to be 20-30% of the spreading factor when the voice activity factor is not used. To support a single air interface for both multi- and isolated-cell-environment DS- CDMA, the link capacity needs to be increased for isolated cells. Since interference from other users and multipath interference make it difficult to achieve orthogonalization in the code domain when the bandwidth is broadened, orthogonalization between simultaneous users in the frequency or time domain must be established. Figure 8 shows the concept called variable spreading and chip repetition factors CDMA (VSCRF- CDMA) [13]. In a multi-cell environment, it is usual to perform spreading only in the time domain, which makes one-cell frequency reuse easy to achieve. In an isolated-cell environment, on the other hand, the principle of symbol repetition is applied to the spread chip sequence and chip repetition is performed. Here, Vol. 2 No. 9 Sep

15 1 symbol = SF cellular chip 1 symbol = SF hotspot chip repetition CRF times Time Time User Time/frequency conversion Frequency Time/frequency conversion Frequency Cellular system Variable spreading factor, chip repetition factor (SF, CRF) Fig. 8. Concept of VSCRF-CDMA. Hotspot indoor cell SF denotes the spreading factor for total band broadening. Its value is determined by the symbol rate of the physical channel. In an isolated-cell environment, CRF is set to 1 or greater, and the time-domain spreading factor SF hotspot is made small based on the relationship SF = SF hotspot CRF. This kind of control makes it possible to allocate received signals from a number of simultaneous users equal to CRF to a set of subcarriers that are mutually orthogonal in the frequency domain. In the method proposed here, CRF and SF hotspot are varied adaptively according to the cell configuration (multiple cells or an isolated cell), number of simultaneously accessed channels, and propagation condition (number of multipaths). 4.5 Physical channel configuration Figure 9 shows the configuration of physical channels. First, the common control channel transmits broadcast and paging information at a fixed level of transmission power. This channel uses a fixed modulation scheme, quadrature phase shift keying (QPSK), and a low coding rate so that reception can be performed at the required quality and coverage within the cell. Next, the shared packet channel transmits high-speed packet data at a fixed level of transmission power. It applies AMC using a modulation scheme and channel coding rate appropriate for the received SIR and provides a maximum throughput that guarantees the required received-packet error rate. Finally, the associated control channel sends control information for the physical and datalink layers to facilitate high-quality transmission on the shared packet channel. It features a fixed modulation scheme (QPSK) and coding rate and applies transmission-power control to compensate for fluctuations in the received level caused by instantaneous fading. 5. Efficient packet access techniques Table 3 lists realtime (RT) and non-realtime (NRT) traffic requirements [14], [15]. The transmission of RT traffic data such as audio and video broadcasts must provide guaranteed reception quality at or below the required residual packet error rate while meeting delay requirements. In contrast, NRT traffic data such as file transfer and WWW browsing must ensure the delay not more than a few seconds, which is a less stringent requirement than that of RT traffic data, but the data transmission must be error free within this delay. Figure 10 shows packet control on the data-link and physical layers [16]. On these layers, scheduling is performed to satisfy the required delay and IP packet loss rate on the RAN in accordance with the traffic data in question. On the downlink, efficient transmission-slot allocation (scheduling) is performed according to the received SIR and type of traffic data (RT or NRT) of each mobile ter- 26 NTT Technical Review

16 Good Received SIR Bad Common control channel: Fixed transmission power and fixed modulation scheme Fixed transmission power Shared/common packet channel: Fixed transmission power with AMC Fixed transmission power Associated control channel: Transmission-power control and fixed modulation scheme Information bit rate Transmission-power control Time Low speed High speed Fig. 9. Physical channel configuration. Table 3. Traffic requirements. Servise class Realtime (RT) Non-realtime (NRT) Example Audio and video broadcasts File transfer, WWW browsing QoS requirements (examples) End-to-end delay Pacekt loss rate <150 ms 0.5% or less <2-3 s 0% minal. On the uplink, each terminal sends a reservation packet beforehand to report the QoS requirements of subsequent data packets, data size, channel conditions, etc., to the base station and the base station performs transmission-slot allocation on the data packet channel for each mobile terminal based on this information. The system also applies MAC-layer ARQ to the packet data channel according to the required delay time. The above approach can reduce the required transmission power through a time diversity effect, especially in the case of NRT traffic data [17]. 6. Multi-antenna transmission/reception techniques 6.1 Configurations A broadband wireless access system is expected to use a high carrier frequency to support high-speed transmission bit rates, and if such a service is to be provided to a wide area, small zones (micro-cells) and low required received E b /N 0 will be indispensable. Therefore, it would be useful to have an adaptive antenna array that can generate directional beams in the angular direction of each user. At the same time, a multiple-input-multiple-output (MIMO), spatially multiplexed multi-antenna transmitting/receiving scheme is effective for increasing the throughput (frequency efficiency). This scheme uses multiple transmitting/receiving antennas and radio devices and transmits different data streams from each transmitter. On the downlink, VSF-OFCDM wireless access with a frequency bandwidth of 100 MHz enables a peak throughput of 200/300 Mbit/s to be achieved by combining 16QAM/64QAM data modulation and a channel coding rate of 3/4. A four-antenna MIMO, for example, might achieve 1-Gbit/s transmission (frequency efficiency: 10 bit/s/hz) for a frequency bandwidth of 100 MHz in an isolated-cell environment. Below we discuss three multi-antenna transmitter/receiver configurations from the viewpoint of improving frequency efficiency. Table 4 Vol. 2 No. 9 Sep

17 Traffic data (User #1) Traffic data (User #K) Detect QoS class (delay, IP packet loss rate) Detect QoS class (delay, IP packet loss rate) Determine priority Determine priority Feed back channel quality information Perform high-speed packet scheduling Allocate transmission slots to each user Fig. 10. Packet scheduling on the data-link and physical layers. Table 4. Comparison of multi-antenna transmitter/receiver configurations. Effect Required transmission-antenna interval Inter-antenna fading correlation RF circuit calibration Large Small Own-cell/other-cell interference-suppression effect MIMO multiplexing MIMO diversity Increases information bit rate Increases diversity gain Large Small signal-separation ability Small transmission-diversity gain Large signal-separation ability Large transmission-diversity gain No Unnecessary Adaptive-antenna-array transmission Increases average received SIR Small Large antenna gain Small antenna gain Yes Necessary summarizes the comparison. MIMO multiplexing method: transmits different data streams on independent radio propagation paths using the same frequency band and time slot [18] MIMO diversity method: performs space-time coding and transmits data streams amongst multiple antennas [19] Adaptive-antenna-array transmission method: performs directional transmission using multiple transmitting antennas (1) The MIMO transmitter/receiver configuration (MIMO multiplexing method) is shown in Fig. 11(a). This method performs serial/parallel conversion on a modulated data stream to produce N separate streams and transmits them by spatial parallelism. Since different data streams are transmitted on the same frequency band and time slot, the received signals carrying these N data streams must be separated at the receiver. For this purpose, the following methods have been proposed. Separate signals by maximum likelihood detection (MLD) Combine received signals by weighting so as to minimize the mean square error Successively extract regenerated data replicas from the received signal (Vertical Bell Laboratories layered space time (V-BLAST)) [20] (2) In the MIMO diversity method, the transmitter performs space time block coding (STBC) [19] after information bits have been channel encoded and data modulated and generates and transmits N streams of coded data. The receiver performs STBC decoding at each antenna and then performs antenna-diversity reception through maxi- 28 NTT Technical Review

18 Transmitter Arrangement of N separated antennas Arrangement of M separated antennas Receiver Transmission data Serial-to-parallel conversion of among N antennas Data modulation section 1 Data modulation section 2 Data modulation section N Transmits different data by different antennas Received-signal separation section Data demodulation section 1 Data demodulation section 2 Data demodulation section N Combines N data streams Increases information bit rate by N times (a) MIMO transmission/reception Received data Transmitter Transmission data Data modulation section Make N-antennas worth of copies Antenna weightgeneration section Close arrangement of N antennas Directional transmission Increases received-signal power by N times Data demodulation section Receiver Received data (b) Adaptive antenna array transmission Fig. 11. Multi-antenna transmission/reception methods. mal ratio combining (MRC). (3) A transmitter/receiver configuration for performing adaptive-antenna-array transmission is shown in Fig. 11(b). In this configuration, the system subjects a data stream to channel coding and modulation (including data modulation and spread modulation), copies the result N times corresponding to the number of antennas, and multiplies each stream by an antenna weight unique to each antenna branch. The antenna weights are generated adaptively by first estimating the direction of a signal received from the target user on the uplink and then computing them so that the main beam of the transmitting antenna pattern is directed toward the user [21]. On the receiving side, the system performs antenna diversity reception by MRC. This type of adaptive-antenna-array transmission can ideally increase antenna gain by N times, where N is the number of transmitting antennas. 6.2 Comparison of configurations for ultrahighspeed signal transmission As shown in Table 4, the MIMO multiplexing method, which transmits information in parallel via multiple transmitting antennas, can increase the information bit rate by simply increasing the number of transmitting antennas, unlike the other two methods. In MIMO diversity and adaptive-antenna-array transmission, the throughput can be increased by increasing either the modulation level or channelcoding rate. For VSF-OFCDM wireless access with a frequency bandwidth of 100 MHz, assuming four antennas for transmission and four for reception, the MIMO multiplexing method can achieve a highspeed, high-efficiency throughput of 1 Gbit/s (10 bit/s/hz) using 16QAM or 64QAM data modulation. The MIMO diversity or adaptive-antenna-array transmission methods, on the other hand, would have to employ high-efficient multilevel modulation, so the achievable throughput performance is degraded due to the decrease in the minimum Euclidean distance. Overall, the MIMO multiplexing method, which transmits multiple data streams by spatial multiplexing, is the most appropriate method for achieving high-efficiency transmission such as 10 bit/s/hz [22]. Vol. 2 No. 9 Sep

19 7. Conclusion This article gave an overview of broadband wireless access technology featuring seamless support of various communication environments from cellular systems to isolated cells like hotspots and offices all by a single wireless interface. In future research, we plan to perform indoor and outdoor experiments to measure propagation characteristics and evaluate packet-signal transmission on a broadband channel. References [1] 3GPP, TR25.848, Physical Layer Aspects of UTRA High Speed Downlink Packet Access. (UTRA: UMTS (universal mobile telecommunications system) terrestrial radio access) [2] M. Tanno, H. Atarashi, K. Higuchi, and M. Sawahashi, Fast Cell Search Algorithm for System with Coexisting Cellular and Hot-spot Cells for OFCDM Forward Link Broadband Wireless Access, IEEE VTC 2003-Spring, pp , Apr [3] Y. Ishii, K. Higuchi, and M. Sawahashi, Three-step Cell Search Algorithm Employing Synchronization and Common Pilot Channels for OFCDM Broadband Wireless Access, IEICE Trans. Commun., Vol. E85-B, No. 12, pp , Dec [4] M. Tanno, H. Atarashi, K. Higuchi, and M. Sawahashi, Three-step Cell Search Algorithm Employing Common Pilot Channel for OFCDM Broadband Wireless Access, IEICE Trans. Commun., Vol. E86-B, No. 1, pp , Jan [5] A. Morimoto, S. Abeta, and M. Sawahashi, Cell Selection Based on Shadowing Variation for Forward Link Broadband OFCDM Packet Wireless Access, IEEE VTC 2002-Fall, pp , Sep [6] N. Yee, J. P. Linnartz, and G. Fettweis, Multi-Carrier CDMA in Indoor Wireless Radio Networks, PIMRC 93, pp , Sep [7] K. Fazel and L. Papke, On the Performance of Convolutional-coded CDMA/OFDM for Mobile Communication Systems, PIMRC 93, pp , Sep [8] S. Abeta, H. Atarashi, M. Sawahashi, and F. Adachi, Performance of Coherent Multi-Carrier/DS-CDMA and MC-CDMA for Broadband Packet Wireless Access, IEICE Trans. Commun., Vol. E84-B, No. 3, pp , Mar [9] H. Atarashi, S. Abeta, and M. Sawahashi, Broadband Packet Wireless Access Appropriate for High-speed and High-capacity Throughput, IEEE VTC2001-Spring, pp , May [10] M. Sawahashi and H. Atarashi, Broadband TD-OFCDM Packet Transmission using Variable Diffusion Rate, 2001 IEICE General Conference, B-5-97, Mar (in Japanese). [11] H. Atarashi, S. Abeta, and M. Sawahashi, Variable Spreading Factor Orthogonal Frequency and Code Division Multiplexing (VSF- OFCDM) for Broadband Packet Wireless Access, IEICE Trans. Commun., Vol. E86-B, No. 1, pp , Jan [12] S. Suwa, H. Atarashi, S. Aheta, and M. Sawahashi, Optimum Bandwidth per Sub-carrier of Multicarrier/DS-CDMA for Broadband Packet Wireless Access in Reverse Link, IEICE Trans. Fundamentals, Vol. E85-A, No. 7, pp , July [13] H. Atarashi, T. Kawamura, and M. Sawahashi, Variable-diffusionrate Multi-carrrier/DS-CDMA Uplink Broadband Wireless Access using Symbol Repetition, 2003 IEICE General Conference, B-5-53, Mar (in Japanese). [14] Recommendation G.114, Telecommunication Standardization Sector of ITU, One-way Transmission Time, Geneva, Switzerland, Mar [15] Recommendation G.1010, Telecommunication Standardization Sector of ITU, End-user Multimedia QoS Categories, Nov [16] A. Harada, S. Abeta, and M. Sawahashi, Adaptive Radio Parameter Control Considering QoS for Forward Link OFCDM Wireless Access, IEICE Trans. Commun., Vol. E86-B, No. 1, pp , Jan [17] Y. Iizuka, M. Tanno, and M. Sawahashi, Efficient random access channel transmission method utilizing soft-combining of retransmitted message packets according to QoS, IEEE 1CCS2002, pp , Nov [18] G. J. Foschini, Jr., Layered space-time architecture for wireless communication in a fading environment when using multi-element antennas, Bell Labs Tech. J., pp , Autumn [19] V. Tarokh, H. Jafarkhani, and R. Calderbank, Space-Time Block Coding for Wireless Communications: Performance Results, IEEE J. Select Areas Commun., Vol. 17, No. 3, pp , Mar [20] P. W. Wolniansky, G. J. Foschini, G. D. Golden, and R. A. Valenzuela, V-BLAST: an architecture for realizing very high data rates over the rich-scattering wireless channel, 1998 URSI International Symposium on Signals, Systems and Computers, pp , Sep [21] H. Taoka, S. Tanaka, T. Ihara, and M. Sawahashi, Adaptive Antenna Array Transmit Diversity in FDD Forward Link for W-CDMA and Broadband Packet Wireless Access, IEEE Wireless Communications, Vol. 9, pp , Apr [22] J. Kawamoto, T. Asai, K. Higuchi, M. Sawahashi, Comparison of Signal Transmission Methods using MIMO Channel of VSF- OFCDM toward 1-Gbit/s Data Transfer, IEICE Technical Report, RCS , Apr (in Japanese). 30 NTT Technical Review

20 Mamoru Sawahashi Director, IP Radio Network Development Department, NTT DoCoMo, Inc. He received the B.S. and M.S. degrees from the University of Tokyo, Tokyo in 1983 and 1985, respectively, and the Dr. Eng. degree from the Nara Institute of Technology in In 1985, he joined NTT Electrical Communications Laboratories. In 1992, he transferred to NTT Mobile Communications Network, Inc. (now NTT DoCoMo, Inc.). He is a member of IEEE and the Institute of Electronics, Information and Communication Engineers of Japan (IEICE). He is currently serving as an editor for IEEE Transactions on Wireless Communications. Kenichi Higuchi Assistant Manager, IP Radio Network Development Department, NTT DoCoMo, Inc. He received the B.S. degree from Waseda University, Tokyo in 1994 and the Ph.D. degree in electrical and communication engineering from Tohoku University, Sendai, Miyagi in In 1994, he joined NTT Mobile Communications Network, Inc. (now, NTT DoCoMo, Inc.). He is a member of IEICE. Sadayuki Abeta Manager, IP Radio Network Development Department, NTT DoCoMo, Inc. He received the B.E., M.E., and Ph.D. degrees in electrical communication engineering from Osaka University, Suita, Osaka in 1993, 1995, and 1997, respectively. In 1997, he joined NTT Mobile Communications Network, Inc. (now NTT DoCoMo, Inc.). He is a member of IEEE and IEICE. Motohiro Tanno Assistant Manager, IP Radio Network Development Department, NTT DoCoMo, Inc. He received the B.E. and M.E. degrees in electrical and electronic engineering from Kyoto University, Kyoto in 1993 and 1995, respectively. In 1995, he joined NTT Mobile Communications Network, Inc. (now NTT DoCoMo, Inc.). He is a member of IEICE. Hiroyuki Atarashi Assistant Manager, IP Radio Network Development Department, NTT DoCoMo, Inc. He received the B.E., M.E., and Ph.D. degrees in electrical engineering from Keio University, Tokyo in 1994, 1996, and 1999, respectively. In 1999, he joined NTT Mobile Communications Network, Inc. (now NTT DoCoMo, Inc.). He is a member of IEEE and IEICE. Taisuke Ihara Manager, IP Radio Network Development Department, NTT DoCoMo, Inc. He received the B.S. degree from the Science University of Tokyo, Tokyo in 1992 and the M.S. and Ph.D. degrees from Tohoku University, Sendai, Miyagi in 1994 and 2003, respectively. In 1994, he joined NTT Mobile Communications Network, Inc. (now NTT DoCoMo, Inc.). He is a member of IEICE. Vol. 2 No. 9 Sep

21 Wireless QoS Control Technologies Takeshi Yamashita, Hidetoshi Kayama, Lan Chen, Daisuke Kitazawa, Masayuki Motegi, and Narumi Umeda Abstract In the fourth-generation (4G) mobile communication system, all communications including realtime communications is expected to be carried by packets multiplexed on a shared radio channel to improve the radio resource efficiency. Because the delay characteristics and the throughput of a packet radio channel vary greatly with changes in both traffic volume and radio channel quality, NTT DoCoMo proposes novel wireless QoS (quality of service) control technologies to guarantee QoS requirements and achieve highly efficient realtime communication with a wide range of transmission rates. 1. Introduction NTT DoCoMo, Inc. Yokosuka-shi, Japan t_yamasita@nttdocomo.co.jp The fourth-generation (4G) mobile communication system is expected to provide a wide variety of applications and services through high-data-rate radio channels. All communications including realtime communications such as VoIP, videophones, video streaming, and digital TV broadcasting, will be packet multiplexed on a shared radio channel to improve the radio resource efficiency. These communication services have individual requirements for bandwidth and acceptable transmission delay, i.e., quality of service (QoS). However, packet radio systems suffer from drastic changes in the transmission delay and channel throughput when packet traffic and radio channel conditions fluctuate. To handle these dynamic conditions, a key technology for implementing the system is wireless QoS control for efficient wireless packet multiplexing that takes into account variations in radio propagation and interference while satisfying the QoS requirements of various applications [1], [2]. This article describes four novel wireless QoS control technologies suitable for high-speed packet communication systems: radio-condition-aware admission control, multistage hybrid scheduling, reservationbased reverse channel access protocol, and adaptive battery conservation management. We are implementing these technologies in a prototype 4G system. 2. Features of wireless QoS control Wireless QoS control is designed to provide the following features. (1) Guaranteed and best-effort services Realtime communications has strict requirements regarding the bandwidth and acceptable transmission delay. To cope with these applications, wireless QoS control provides guaranteed services that can provide a flexible guaranteed bandwidth and delay suitable for realtime communications. Best-effort services will also be provided for non-realtime communications and these services will take into account relative packet priorities with/without radio channel conditions. (2) Smooth working with IP-QoS By assigning guaranteed-class packets to the expedited forwarding (EF) class for the differentiated services code point (DSCP) and best-effortclass packets to the best effort (BE) class for the DSCP, we ensure that wireless QoS control can work smoothly with Internet protocol (IP) QoS. (3) Area-free guaranteed services To improve radio channel efficiency, the adaptive modulation and coding (AMC) scheme is 32 NTT Technical Review

22 Table 1. Examples of services provided according to wireless QoS control. Technologies Service/applied technology Multistage hybrid scheduling Services Guaranteed QoS Target application Admission control Scheduling between guaranteed and best-effort classes Scheduling between same-class users Guaranteed-class services Guaranteed-rate service is provided regardless of variation in interference and user location. Guaranteed rate is set flexibly considering a variable rate and the user maximum and minimum requirements. Transmission rate and delay Realtime application Applied (QoS request and radio conditions are considered.) Guaranteed rate, admitted delay, fairness, etc. are considered. Best-effort service Best-effort-class services Non-realtime application Not applied Priority control is applied in the order: packets within the guaranteed rate, best-effort packets, and packets that exceed the guaranteed rate. Fairness and radio conditions are considered. assumed to be used in 4G. This scheme changes the transmission rate according to the signal to interference-plus-noise ratio (SINR). However, from the user s viewpoint, the rate should be constant regardless of his/her location. Our radiocondition-aware admission control enables stable and guaranteed services with a flexible guaranteed rate to be provided to realtime users regardless of the variation in interference and the location of the mobile station. Examples of services that can be provided depending on the type of wireless QoS control used are shown in Table Architecture of packet transmission using wireless QoS control Wireless QoS control aims to provide a range of guaranteed-rate and area-free realtime communication services regardless of the packet traffic and radio condition fluctuations. Smooth working with IP-QoS and channel efficiency are also taken into account. Based on this concept, we propose a novel wireless QoS control [3]. The architecture for packet transmission is shown in Fig. 1. It comprises several control factors for the medium access control (MAC) and radio resource control (RRC) layers and a radio resource control factor over these two layers. These factors can work smoothly with IP-QoS. IP packets are first mapped to the best-effort or guaranteed classes. For guaranteed-class users, radio-condition-aware admission control is carried out to avoid an overload among these users. For best-effort-class users, classification is carried out according to the DSCP implying their relative priorities. In multistage hybrid scheduling, guaranteed scheduling performs traffic shaping to guarantee both the admitted rate and delay requirements of guaranteed-class users, while besteffort scheduling considers radio resource efficiency and fairness among users and classes. In addition, multistage scheduling gives priority to guaranteedclass users. Transmission parameters are determined according to the QoS class and channel conditions based on information sent from the receiver. In the downlink, radio resources can be allocated in a straightforward scheme. In the uplink (reverse link), a reservation-based access protocol is employed to enable priority-based resource allocation required for the wireless QoS control technology (See section 4.). 3.1 Radio-condition-aware admission control (RAC) These days, many communication applications have functions for responding flexibly to changes in transmission rates by using variable rate coding. Therefore, the guaranteed rate can be set flexibly within a variable rate range, that is the maximum and minimum requirements. This flexible admission control allows more users to be accommodated. Making allowances for user movement and variations in interference, the lowest modulation coding set is used as the channel capacity criterion for admission control decision, as indicated in Fig. 2(a), which shows the transmission rate for three different modulation coding sets (MCSs). To utilize the statistical multiplexing characteristics of packet communications for radio resource efficiency, the resource utilization rate of the guaranteed-class users is also considered. However, Vol. 2 No. 9 Sep

23 Diffserv IP Guaranteed-class packets Mapping Best-effort-class packets Radio resource management Radio-condition-aware admission control Guaranteed scheduling Multistage hybrid scheduling Multistage scheduling Transmission parameter selection Classification Best-effort scheduling RRC MAC Resource allocation Packet transmission Fig. 1. Architecture of packet transmission. Transmission rate High-rate MCS Realtime measured traffic (x Mbit/s) Acceptable rate Reference capacity Low-rate MCS (y Mbit/s) SINR (a) Variation in transmission rate in adaptive modulation Margin (α) Reference capacity (y Mbit/s) (b) Acceptable rate Fig. 2. Principle of radio-condition-aware admission control. not all the guaranteed-class users encounter poor conditions. To improve the resource utilization efficiency, the measured resource utilization rate for guaranteed-class users is considered, as shown in Fig. 2(b). Here, the total of the bandwidth requests that can be accepted (i.e., the acceptable rate) is given by the reference capacity (y) minus the measured resource utilization rate (i.e., the realtime measured traffic (x)) and a margin (α). That is, the acceptable rate = y - (x + α). 3.2 Multistage hybrid scheduling (MHS) The scheduling that determines the transmission order when all user packets are multiplexed onto one common radio channel is important for maintaining service quality and supporting various applications and service classes. Furthermore, in wireless communication systems, an important purpose of scheduling is to increase the efficiency of radio resource utilization. There have been many proposals for scheduling that takes into account QoS requests [4] or that aims to increase the radio resource utilization efficiency [5]. However, in the 4G system, there are demands to provide a range of guaranteed rates and maintain the service quality for realtime communications without deteriorating the radio resource utilization efficiency. 34 NTT Technical Review

24 EF Guaranteed class BE Best-effort class 1 2 n n + 1 n + 2 n Buffer per user Buffer per user or class Guaranteed scheduling Best-effort scheduling Multistage scheduling Fig. 3. Principle of multistage hybrid scheduling. To satisfy these demands, multistage hybrid scheduling (MHS) has been proposed [6], [7]. Its principle is shown in Fig. 3. Here, we explain the downlink transmission case. As described above, each packet is classified as either a guaranteed- or best-effort-class packet; that is, IP packets with a header of EF and BE are mapped into the guaranteed and best-effort classes, respectively. Either each user has his/her own individual buffer or best-effort-class packets with the same DSCP can share a buffer because, from the viewpoint of the QoS requirement, flow control based on the guaranteed rate and acceptable delay are necessary for individual guaranteed-class users, while class-based flow control is sufficient for best-effortclass users. However, for scheduling that takes into account the radio resource utilization efficiency, it is necessary to consider the radio propagation conditions of each user, and individual buffers are required not only for the guaranteed-class users, but also for best-effort-class users. In MHS, a different scheduling scheme is used between the guaranteed- and best-effort-class users. Guaranteed scheduling considers guaranteeing both the admitted rate and delay requirements of guaranteed-class users, while best-effort scheduling considers the radio resource efficiency and fairness among all best-effort-class users. Examples of scheduling applicable to both types of users are described below. (1) Guaranteed class Weighted round robin (WRR): Each user in turn sends a weighted number of packets. Header early first (HEF) [8]: A time stamp is attached to each packet. The time stamps of the packets at the tops of the buffers are compared and the packet with the earliest time stamp is scheduled first. In both of these methods, the number of packets transmitted within one turn depends on the guaranteed rate declared by the users. (2) Best-effort class Proportional fairness (PF) [9]: The ratio of instantaneous SINR to averaged SINR is measured for each user. The packets are scheduled based on this ratio in order from highest to lowest. This method takes into account the radio resource efficiency and fairness among users. Class-based queuing (CBQ): When users of the same DSCP class share a buffer and DSCP class-based QoS control is used, weights are assigned to each class according to its priority and packets are taken from the buffers according to their weights. (3) Scheduling between guaranteed and best-effort classes (multistage scheduling) In multistage scheduling, guaranteed-class packets within the guaranteed rate are selected first. Then, packets of best-effort-class users are selected for transmission. Finally, packets that exceed the guaranteed rate are selected. Resources are assigned on a block basis, and in the case of a guaranteed-class user, the guaranteed rate is also considered during resource allocation. Here, the transmission block size of a user is determined according to his/her radio channel conditions using a channel quality indicator during transmission. Vol. 2 No. 9 Sep

25 3.3 Performance of RAC and MHS Computer simulation results for the satisfaction rate for guaranteed-class users and the transmission rate for best-effort-class users are shown in Fig. 4. In these figures, round robin (RR) indicates the performance of a method without RAC and MHS, while priority scheduling (PS) indicates the performance of a method that gives unconditional priority to the guaranteed-class users. We can see that wireless QoS can increase the satisfaction rate of guaranteed-class users without greatly degrading the transmission rate of best-effort-class users by setting the guaranteed rate dynamically within the range between the maximum and minimum rates requested by users and by performing scheduling that takes into consideration the guaranteed rates. RAC and MHS make it possible to aim for efficient utilization of the radio resources while guaranteeing the delay and transmission rate for guaranteed-class users. These methods can simultaneously maintain some degree of fairness among best-effort-class users by taking into account the radio resource efficiency and fairness [3], [6]-[8]. 4. Reservation-based reverse channel access protocol [10] For downlink packet transmission, resource allocation on a packet-by-packet basis is relatively easy because the base station can manage all the transmissions on the downlink packet channel. On the other hand, packets in the reverse channel are generally transmitted in a random access manner. In this manner, a massive transmission with the same timing causes packet loss due to multiple access interference. As a result, each realtime packet may suffer an unacceptable transmission delay caused by a random access retry under relatively heavy traffic conditions. According to the wireless QoS architecture (See section 3.), all packets are categorized into two classes: guaranteed and best-effort. Realtime packets are categorized into a guaranteed class, in which the transmission rate and delay are guaranteed by RAC and MHS. Incidentally, in MHS, the guaranteed-class packets that are below the guaranteed rate are always transmitted before best-effort-class packets. In the downlink channel, the transmission order of the packets can be easily arranged because all packets are assembled at a base station and arranged by centralized control. On the other hand, packets in the reverse channel are not; they are accumulated in distributed queues from individual mobile stations. To satisfy the QoS requirement for realtime communications and to increase the radio resource utilization efficiency in such a distributed queue environment, two schemes have been proposed. (1) Prioritized resource allocation (PRA) [11] This is a reservation-based scheme. Two different resource allocation areas in the time domain are established for the guaranteed and best-effort classes. The guaranteed-class packets have a wider allocation area so that they can obtain radio resources before the best-effort-class packets can. With this scheme, almost all the bandwidth can be occupied by guaranteed-class packets. Figure 5 shows an example of the time slot allocation in the PRA scheme. (2) Reservation-based random access protocol with a temporarily assigned dedicated control channel Satisfaction rate (%) 90 Transmission rate (kbit/s) Wireless RR QoS 40 Wireless 50 QoS 40 PS 20 PS Rv = 0.5 RR Rv = Number of users per cell Number of users per cell (a) Satisfaction rate for guaranteed-class users (b) Transmission rate for best-effort-class users 100 Fig. 4. Performance of RAC and MHS. Rv = 0.5 means 50% of all users are guaranteed-class users. 36 NTT Technical Review

26 Guaranteed class Best-effort class (Allocation not possible) (Allocation possible) Allocated Time Time slots allocatable to best-effort-class packets Time slots allocatable to guaranteed-class packets Initial allocatable pointer Fig. 5. Example of time slot allocation of PRA. (RAP-TDC) [12], [13] In this protocol, only a low-rate temporarily dedicated control channel (TDCCH) is temporarily assigned between a base station and a mobile station during a packet burst. In this channel, reservation signals for access and transmission control parameters are transmitted simultaneously. After a reservation has been accepted, user packets are transmitted on assigned common packet channel (CPCH) slots. This scheme prevents the packet-by-packet setup delay by maintaining transmission parameters using the TDCCH. In addition, the mobile station can transmit its reservation signal without any conflict with other reservation signals, so this scheme enables the short delay packet transmission that is required for realtime communications. The interleave reservation mechanism that enables us to remove the channel inefficiency caused by the round trip delay of the reservation protocols is also employed. In addition, because the reservation signal is transmitted without collisions, when the reservation is denied because of a temporary lack of resources, the terminal can continuously retry the request. Transmission control signals, such as the transmission power control (TPC) command, channel quality indicator, and acknowledgment signals, are continuously exchanged in the TDCCH to enable immediate packet (re)transmissions in both directions. If no packets are transmitted in the CPCH during a predetermined time, which is referred to as the release delay time below, then the TDCCH is released. It takes a little time to reestablish the TDCCH, so the release delay time should be adaptively set according to the QoS requirement of the transmitted packets. In addition, it is effective to coordinate the release delay time with a dormant control timer such as the timer described in section Adaptive battery conservation management In existing mobile communication systems, a mobile station intermittently receives paging signals that are periodically transmitted by a base station during its stand-by time (which is defined as the idle mode in this paper). This scheme is generally used for battery conservation. Moreover, since the termination of a session can be clearly defined in a connection oriented manner, the mobile station can immediately transit to the idle mode after a session ends. On the other hand, in connectionless communications, session initiation and termination are generally indistinct, so it is effective to employ a timer that is initiated after each packet has been received or transmitted. When this timer expires, it initiates the transition to idle mode. When a packet addressed to a mobile station that is in idle mode reaches a base station, the base station informs the mobile station by sending packet arrival notifications (i.e., by paging it) using a periodically transmitted signal. Therefore, this scheme involves a wake-up delay. In the conventional battery conservation method, because there is only one timer for entering idle mode and one paging interval, the requirement for the acceptable transmission delay of realtime communications cannot be satisfied when applying this method to the 4G system. To overcome these problems, adaptive battery con- Vol. 2 No. 9 Sep

27 Active mode BSM-EF Battery saving mode BSM-BE Idle mode State in which the control signal from the base station is received intermittently Fig. 6. State transitions in adaptive battery conservation management. servation management (ABCM), which takes into consideration the QoS requirements, has been proposed for connectionless communication systems such as the 4G system [14]. In packet communications, the communication state of the mobile station can be classified into three states where i) packets are transmitted and received continuously, ii) no packets are transmitted or received for a long time, and iii) no packets are transmitted or received for a short time. ABCM pays attention to this feature, and a new state called battery saving mode is defined to improve battery conservation. The ABCM state transition diagram for the mobile station is shown in Fig. 6. Three mobile station states are defined: the active, idle, and battery saving modes. For the battery saving mode, two submodes (BSM-EF and BSM-BE) are defined that correspond to the two QoS classes. The guaranteed-class mobile station transits to BSM-EF and the best-effort-class mobile station transits to BSM-BE. The transition to the corresponding state is triggered by the transmission and reception of packets or by the expiration of a timer. The battery saving mode is an intermediate state between the active and idle modes, and the control signal from the base station is received intermittently, as in the idle mode. The difference between the battery saving and idle modes is that the intermittent receiving cycle is shorter in the former, and BSM-EF is assigned a shorter period than BSM-BE because it has stricter delay requirements. Furthermore, the timer for the transition from active mode to battery saving mode is also set corresponding to the guaranteed and best-effort classes, so the timer for the guaranteed class is longer than that for the best-effort class. This method makes it possible to guarantee the packet transmission delay for the guaranteed class, although battery conservation is equivalent to that of the conventional methods without an intermediate state like the battery saving mode. For the best-effort class, on the other hand, effective battery conservation is possible (although an increase in delay equivalent to at most the intermittent receiving cycle is unavoidable). Simulation results indicate that ABCM achieves better battery conservation than conventional methods. 6. Conclusion This article described some wireless QoS control technologies that have been proposed for the 4G system and are currently under evaluation. In the future, we plan to evaluate them in more detail by simulation and test their compatibility and effectiveness for various applications in a radio environment using a prototype system. References [1] Y. Yamao, N. Umeda, T. Otsu, and N. Nakajima, Fourth Generation Mobile Communications System Issues Regarding Radio System Technologies, IEICE Trans., Vol. J83-B, No. 10, Oct [2] S. Ohmori, Y. Yamao, and N. Nakajima, The future generations of mobile communications based on broadband access technologies, IEEE Communications Magazine, pp , Dec [3] L. Chen, H. Kayama, Y. Kawabe, and N. Umeda, Performance of a Novel Wireless QoS Proposed for Broadband CDMA Packet Communication Systems, IEEE Proc. VTC-2004 Spring, [4] P. Ferguson and C. Huston, Quality of Service, John Wiley & Sons, [5] Y. Cao and V. O. K. Li, Scheduling Algorithms in Broad-Band Wireless Networks, Proceedings of IEEE, Vol. 89, No. 1, pp , Jan [6] L. Chen, H. Kayama, and N. Umeda, Wireless QoS Architecture for High Speed CDMA Packet Cellular system, IEICE 2003 Annual 38 NTT Technical Review

28 Conference, B-5-138, [7] L. Chen, H. Kayama, and N. Umeda, Radio Condition Aware Admission Control and Hybrid Scheduling for High Speed CDMA Packet Cellular System, IEICE 2003 Annual Conference, B-5-139, [8] D. Kitazawa, L. Chen, H. Kayama, and N. Umeda, A Study on Packet Scheduling Considering Buffering Delay for High Speed CDMA Cellular System, IEICE 2003 Annual Conference, B-5-140, [9] A. Jalali, R. Padovani, and R. Pankaj, Data Throughput of CDMA- HDR a High Efficiency, High Data Rate Personal Communication Wireless System, IEEE Proc. VTC-2000 Spring, [10] H. Kayama, H. Qiu, and N. Umeda, Prioritized Reservation Type MAC Protocol for Broadband CDMA Packet Communication Systems, IEEE Proc. VTC-2004 Spring, [11] H. Kayama, H. Qiu, and N. Umeda, Prioritized Resource Allocation (PRA) for high-speed uplink packet channel, IEICE 2003 Annual Conference, B-5-145, [12] M. Nagatsuka, Y. Ishikawa, J. Hagiwara, T. Nakamura, and S. Onoe, Capacity evaluation of a Media Access Control method in W- CDMA Packet Mobile Communication, IEICE 1998 Annual Conference, B-5-164, [13] H. Qiu, H. Kayama, and N. Umeda, A Study on Reservation Access Protocol with Dedicated Control Channel (RAP-DC) in High-speed Packet Communication CDMA System, IEICE 2003 Annual Conference, B-5-146, [14] M. Motegi, H. Kayama, and N. Umeda, Adaptive Battery Conservation Management for Multimedia Mobile Packet Communications, IEICE Technical Reports, NS IN , Mar Takeshi Yamashita Research Engineer, Communication Systems Laboratory, Wireless Laboratories, NTT DoCo- Mo, Inc. He received the B.E. and M.E. degrees from Nagaoka University of Technology, Nagaoka, Niigata in 1996 and 1998, respectively. During , he was with NTT DoCoMo, Inc., Radio Network Development Department, where he was involved in the development and study of mobile satellite communication systems. Since 2002, he has been with the Communication Systems Laboratory of the Wireless Laboratories. His research interests include handoff, wireless QoS, and moving networks. He is a member of the Institute of Electronics, Information and Communication Engineers (IEICE). Hidetoshi Kayama Executive senior researcher, Director of Innovative Radio Transmission Laboratories, DoCo- Mo Beijing Communications Laboratories. He received the B.S. and M.S. degrees in electrical engineering and the Ph.D. degree in informatics from Kyoto University, Kyoto in 1987, 1989, and 2004, respectively. In 1989, he joined NTT Radio Communication System Laboratories, where he engaged in R&D of PHS packet data systems, 19-GHz WLANs, and a. From 1998 to 2004, he was with NTT DoCoMo, Inc. where he engaged in R&D of PHS and 4G mobile communication systems. His research interests include access protocols, QoS control, system architecture, and transmission technologies for packet radio systems. He is a member of IEICE and IEEE. He received the scholarship encouragement award from IEICE in 1995 and the best paper award of the International Conference on Telecommunications at ICT2002. Daisuke Kitazawa Corporate Strategy & Planning Department, NTT DoCoMo, Inc. He received the B.S. and M.S. degrees in communication engineering from Osaka University, Suita, Osaka in 1998 and 2000, respectively. At university, he studied mobile communication systems, multi-layer (hierarchical) cellular systems, and software-based multimedia communication systems. He was especially interested in Software Radio and he devoted two years toward his masters degree to improve the system capacity of a software-based mobile communication system using two types of access schemes, TDMA and CDMA. He made two presentations about his research to IEICE. In 2000, he joined the Wireless Laboratories, NTT DoCoMo, Inc. and researched beyond-imt-2000 (4G) mobile communication systems, especially packet scheduling, admission control, and radio resource management. In 2003, he moved to the corporate strategy & planning department. Masayuki Motegi Radio Network System Group, IP Radio Network Development Department, NTT DoCoMo, Inc. He received the M.E. degree in electrical and computer engineering from Yokohama National University, Yokohama, Kanagawa in 2001 and the same year he joined the Wireless Laboratories, NTT DoCoMo, Inc., Yokosuka, Kanagawa. In 2004, he moved to the IP Radio Network Development Department. He is a member of IEICE. Lan Chen Director, Advanced Radio System Lab, Executive senior researcher, DoCoMo Beijing Communications Laboratories. She received the M.S. and Ph.D. degrees in communications and computer engineering of Kyoto University, Kyoto. She joined the Wireless Laboratories of NTT DoCoMo in 1999, where she researched wireless QoS control and radio resource management for 4G and beyond. Her research interests include wireless QoS, radio resource management, packet scheduling, and admission control. She received the best paper award at ICT2002. She is a member of IEEE and IEICE. Narumi Umeda Director, Communication Systems Laboratory, Wireless Laboratories, NTT DoCoMo, Inc. He received the B.E. and M.E. degrees in electronic engineering from Hokkaido University, Sapporo, Hokkaido in 1985 and 1987, respectively. He joined NTT Laboratories in In 1992, he transferred to NTT Mobile Communications Network, Inc. (now NTT DoCoMo, Inc.). He has been engaged in research on the radio link control for the personal digital cellular (PDC) system and IMT His current research interests are in the radio control for 4G mobile communications systems (beyond 3G). He was a co-recipient of the Japan Institute of Invention and Innovation (JIII) Imperial Invention Prize in 1998 and the best paper award at ICT2002. He is a member of IEICE and IEEE. Vol. 2 No. 9 Sep

29 IP-based Mobility Management Technology Ichiro Okajima, Masahiro Inoue, Koji Omae, Hideaki Takahashi, and Narumi Umeda Abstract This article reviews recent progress in the study of IP-based mobility management technology for enhancing the mobility of cellular phones and other mobile nodes in fourth-generation mobile communications systems. The architecture and components of this management technology are described and some experimental and simulation results are presented. 1. Introduction NTT DoCoMo is developing IP-based mobility management technology, which is basic technology needed for the deployment of fourth generation mobile communications systems (4G systems) that will support broadband mobile communications services. IP-based mobility management technology works at the Internet protocol (IP) layer to enable seamless handoff * of mobile nodes (such as cellular phones, personal digital assistants (PDAs), and notebook computers) while they are moving around in the 4G system and to give packets end-to-end reachability to and from mobile nodes. In short, this technology enables true mobility by permitting mobile nodes to move about freely anywhere in the 4G system and initiate and perform communications. This article presents an overview of the architecture of IP-based mobility management and describes its key components: multiple interface management, active state mobility management, and dormant state mobility management. 2. IP-based mobility management architecture This section outlines the key requirements for mobility management in the 4G system and presents NTT DoCoMo, Inc. Yokosuka-shi, Japan okajimai@nttdocomo.co.jp a management architecture that satisfies them. 2.1 Requirements There are three main requirements for mobility management. (1) High packet transmission quality To support a diverse range of applications, good packet transmission quality is essential, including low packet transmission delay, minimal packet transmission delay deviation, and a low packet loss rate. (2) Modest control cost To make efficient use of wireless links, which have significantly lower bandwidth capacity than wired links, it is necessary to reduce the amount of signaling traffic associated with mobility management that must be transferred over the wireless links. (3) Seamless mobility The management scheme must implement seamless mobility that supports mobile nodes connected to not only 4G cellular wireless links but also IEEE and other wireless links and to Ethernet and other wired links. 2.2 Configuration and protocol stack In principle, mobility management can be implemented at any layer from the link layer to the application layer. However, if mobility management is implemented at the link layer, the applicability would * Handoff: The act of transferring a wireless signal from one cell site to another. Equivalent to handover. 40 NTT Technical Review

30 be limited to the same type link, and this would contravene Requirement (3). Moreover, if mobility management is implemented at the transport layer or the application layer, it would have to be implemented separately for each different transport and application layer protocol, which would increase the amount of signaling traffic and thus fail to satisfy Requirement (2). In contrast, if we implement mobility management at the IP layer, which is common to all link layer, transport layer, and application layer protocols, then Requirements (2) and (3) can both be satisfied. This approach can handle the wireless link disconnections associated with handoffs and delays in packet route setup processing that are the primary factors in the degradation of packet communication quality, so Requirement (1) is also satisfied. Consequently, this approach satisfies all three of the basic requirements. Figure 1 shows a schematic representation of our IP-based mobility management scheme and protocol stack, assuming the use of IPv6 (Internet protocol version 6), which has an enormous address space that can support an extremely large number of mobile nodes. Three of the key terms related to mobile nodes shown in Fig. 1 are explained below. Multiple interface management (MIM) provides the link layer that best matches user preferences to the layers above from among the several link layers available for the mobile node while minimizing the power Correspondent node Home agent Internet Paging agent Mobility anchor point (MAP) IP network Router 4G cellular access point Router with built-in 4G cellular access transceiver IEEE access point Mobile router Mobile IP network Mobile node Applications TCP/UDP TCP: transmission control protocol UDP: user datagram protocol Applications TCP/UDP HMIP-Bv6/IPPv6 MIM G MN 4G Link 4G-AP IPv6 Link Link Routers HMIP-Bv6 Link Link MAP IPv6 Link Link Routers IPPv6 Link Link PA IPv6 Link Link Routers MIPv6 Link Link HA IPv6 Link Link Routers IPv6 Link CN Fig. 1. IP-based mobility management architecture and protocol stack. Vol. 2 No. 9 Sep

31 consumption of the mobile node s other link layers that are not used. HMIP-Bv6 (hierarchical mobile IPv6 with buffering at MAP) is a mobility management scheme that uses a mobility anchor point (MAP) with buffering capability for handoffs. It achieves reliable routing of packets to the mobile node in the active state of communication with a low packet loss rate and low packet transmission delay. IPPv6 (IPv6 paging protocol) is a mobility management protocol that uses a paging agent (PA) to manage the location of mobile nodes in the dormant communication state in local areas and to notify the mobile node of the arrival of incoming packets, thus enabling it to have less frequent packet route setup processing. 3. Multiple interface management This section defines the requirements for interface management with multiple kinds of link layers and then presents an overview of NTT DoCoMo s proposed MIM. 3.1 Requirements There are two main requirements for MIM. (1) NIC selection based on user preferences A network interface card (NIC) is the communication interface hardware that implements the link layer protocol. NICs are rapidly coming down in price and levels of chip integration are increasing, so it is only a matter of time before a mobile node can support multiple NICs. When that happens, MIM will need to be able to determine what NICs are available on a mobile node and select the one that best matches the user s preferences. (2) Lower battery consumption To conserve the mobile node s battery power, MIM must efficiently manage the power consumption of the NICs. 3.2 MIM configuration Figure 2 shows an MIM configuration that satisfies the above requirements [1]. It is built into the mobile node. (1) User preference information User preference information is input via an MIM MIM GUI Application protocols TCP/UDP HMIP-Bv6/IPPv6 MIM User preference information Optimum NIC selection function IEEE link IEEE NIC IEEE access point 4G cellular link 4G cellular NIC 4G Cellular access point Ethernet link Ethernet NIC Ethernet hub IP packet flow LAPI signaling flow MIM internal data flow Fig. 2. MIM configuration. 42 NTT Technical Review

32 graphical user interface and stored in a database. Typical information will include the user s preferences for bit rate, transmission quality, and transmission cost. (2) Optimal NIC selection function The MIM periodically checks to make sure that the NIC being used best matches the user preferences and, if necessary, changes NICs. The MIM also helps conserve battery power by turning off the power to all NICs that are currently not selected. This optimal NIC selection function utilizes the link layer application programming interface (LAPI) [2] to control the link layer and exchange status information between the link layer and the higher layer. For example, if the user prefers a high bit rate, this function selects the NIC that provides the highest bit rate. 4. Active state mobility management This section defines the requirements for active state mobility management, outlines NTT DoCoMo s proposed HMIP-Bv6 scheme, and presents some recent evaluation results for this scheme. 4.1 Requirements There are three main requirements for active state mobility management. (1) Mobile node and mobile network mobility Support of standalone mobile nodes is essential, but the mobility of mobile networks such as groups of mobile nodes traveling in a train or bus must also be supported. (2) Ease of deployment To promote rapid and cost-effective deployment of 4G systems, the equipment must be easy to introduce. (3) High packet communication quality To support a broad range of applications, the management scheme must satisfy the requirements in Table 1 for voice and other realtime traffic to minimize degradation of quality during handoffs, while at the same time minimizing the slowdown of throughput during handoffs to accommodate data traffic. 4.2 HMIP-Bv6 scheme In the conventional mobile IPv6 (MIPv6) [3] and hierarchical mobile IPv6 (HMIPv6) [4] schemes, mobility is achieved by a home agent (HA) and a MAP, which forwards packets destined for mobile nodes, but these schemes do not satisfy the above requirements due to bursty packet losses that occur during handoffs. This led us to propose a new approach called hierarchical mobile IPv6 with buffering at MAP (HMIP-Bv6), which adds a number of new capabilities to the HMIPv6 scheme [5], [6]. The four described below are achieved through extensions made only to the mobile node, MAP, and HA. (1) Agent discovery function This function enables a mobile node to dynamically discover a MAP. This enables the mobile node to select the optimal MAP when there are multiple MAPs on the IP network by comparing the relative distances to MAPs, packet processing loads, and other criteria. We define this capability in an agent discovery protocol [7] and can also use it to discover PAs, as mentioned in section 4. Another advantage of this approach is that no data about MAPs and PAs needs to be preconfigured in the mobile nodes or routers, which simplifies the deployment of MAPs and PAs on the IP network. (2) MAP buffering function Packets destined for a mobile node are buffered in the MAP to prevent packet loss during handoffs. This is illustrated for the HMIP-Bv6 scheme in Fig. 3. (3) Subnet prefix management function This function manages all the subnet prefixes in HAs and MAPs used by the mobile network [8] and supports mobility for all nodes on the mobile network. (4) Care-of address fast configuration function This function speeds up the configuration process- Table 1. Voice communication quality requirements and measured results. Evaluation criteria Requirement [12] MIPv6 Simulation Experiment HMIPv6 HMIP-Bv6 Simulation Experiment Simulation Experiment No. of packets lost per handoff Packet loss rate (%) End-to-end packet transmission delay deviation (ms) Average end-to-end packet transmission delay (ms) Failed to meet requirements Vol. 2 No. 9 Sep

33 Mobile node Router 1 MAP HA CN Handoff needs to be done Binding update (with buffering on) Binding acknowledgment Link layer handoff Care-of address configuration Router 2 Packet buffering Binding update (with buffering off) Binding acknowledgment Transmission of data packets that have been buffered Data packets Control packets Fig. 3. HMIP-Bv6 scheme. ing during handoffs for care-of addresses (packet forwarding addresses of mobile nodes), thereby reducing the packet transmission delay after handoffs. 4.3 Evaluation of HMIP-Bv6 scheme The packet transmission quality of three schemes (MIPv6, HMIPv6, and HMIP-Bv6) was evaluated through simulations and experimental trials. This section presents our findings Evaluation method 1) Evaluation criteria Using the network model shown in Fig. 4, we evaluated the quality of voice and data communications at the mobile node when the voice and data were sent from the correspondent node to the mobile node. For VoIP (voice over IP) speech communications, the criteria were the number of packets lost per handoff, packet loss rate, end-to-end packet transmission delay deviation, and average end-to-end packet transmission delay. For data communications, the criterion was the time required to transfer a file. 2) Network model The network model consists of a correspondent node, an HA, and an IP network interconnected by the Internet. A MAP and several routers are deployed on the IP network and a mobile node performs handoffs between routers. Packet transmission delays through the Internet and the IP network were 50 ms [9] and 10 ms [10], respectively. 3) Traffic models We used two traffic models. For voice traffic, the packet rate was 50 packets per second and the packet size was 80 bytes [11]. For data traffic, a 5-Mbyte file was downloaded by FTP (file transfer protocol) Measured results The measured voice communication quality and the requirements are shown in Table 1. The HMIP-Bv6 scheme was able to satisfy the requirements for the number of packets lost per handoff and packet loss rate as a result of the effectiveness of the MAP packet buffering, but the other two schemes failed to meet these requirements. All of the schemes met the requirement for end-to-end packet transmission delay deviation and average end-to-end packet transmission delay. Figure 5 shows the measured data communication quality. Download times were measured for all three schemes for mobile node handoff intervals of 8, 16, and 32 s and without handoff. The experimental measurements and simulation results showed similar tendencies. The downloads took longer as the interval between handoffs decreased for the MIPv6 and 44 NTT Technical Review

34 Home agent Packet transmission delay through the Internet: 50 ms MAP Packet transmission delay through an IP network: 10 ms Correspondent node Router (i.e., active state). (2) Seamless resumption of communication The packet communication quality must be equivalent to the quality requirements of the active state mobility management so that the dormant state mobility management can resume communications with high packet communication quality whenever the mobile node transits from the dormant state to the active state. Download time (s) Mobile node Fig. 4. Network model. Simulation results MIPv6 HMIPv6 HMIP-Bv6 Fig. 5. Data communication quality. HMIPv6 schemes, but stayed more or less constant for the HMIP-Bv6 scheme. The main factor degrading download speed is the loss of the TCP (transmission control protocol) data segment. Since this does not occur in the HMIP-Bv6 scheme, the download time was unaffected by handoff. 5. Dormant state mobility management This section defines the requirements for dormant state mobility management and overviews and evaluates NTT DoCoMo s proposed IPPv Requirements There are two main requirements for dormant state mobility management. (1) Reduction of control signals Mobility management must be implemented in such a way that the volume of control signals is substantially less when the mobile node has no packets to send or receive (i.e., dormant state) than when the mobile node does have packets to send or receive 5.2 IPPv6 configuration IPPv6 has been proposed as a dormant state mobility management scheme that satisfies the above requirements [13]. This protocol has five functions, which are implemented in mobile nodes and PAs. 1) PA discovery function The mobile node can discover PAs. It dynamically discovers the nearest PA using the agent discovery protocol. 2) Dormant state detection function Figure 6(a) shows the transition of the mobile node to the dormant state. By monitoring how long a mobile node is in a continuous state of non-communication, IPPv6 can determine transitions to the dormant state and send an area registration request to the PA. 3) Paging area formation function Figure 6(a) also shows the mobile node s location management in terms of paging area units. Since there is no need to form a new paging area as long as the mobile node moves within the same paging area, the volume of control signals is significantly less than with active state mobility management. 4) Paging function Figure 6(b) shows how paging request messages are sent throughout the paging area of a mobile node when packets destined for the mobile node are received. It also shows the transition of the mobile node to the active state when the mobile node receives the paging request. 5) Mobile node address packet buffering function The PA buffers packets destined for a mobile node to prevent packet loss while the mobile node is being notified of incoming packets by the paging function. Experimental results Without handoff 32-s intervals 16-s intervals 8-s intervals Vol. 2 No. 9 Sep

35 PA Internet IP network w Area Router registration request 4G-AP AP Mobile node q Non-communication state e Transition between access points (sending not required) AP: access point r Transition between subnets (sending not required) (a) Dormant state transition y Area registration request 4G-AP t Transition between paging areas q Incoming packets w Packet buffering e IP paging 4G-AP HA PA Internet IP network r IP paging response t Packet routing Mobile node Router AP (b) Paging function 4G-AP Router Paging area Fig. 6. IPPv6 operation. 5.3 Evaluation of the IPPv6 scheme The effectiveness of IPPv6 at reducing the number of control packets was evaluated through simulation. Figure 7 shows the number of control packets versus the mobile node handoff interval for four different numbers of subnets per paging area. The number of control packets was normalized by the number of control packets in active state mobility management. The traffic session of mobile nodes had an average interval of 1800 s. The shorter the handoff interval of mobile nodes was (i.e., the more frequently the mobile nodes moved), the more effective IPPv6 was at reducing the number of control packets. This was true for all numbers of subnets. When the paging area consisted of seven subnets and the average handoff interval was 20 s, IPPv6 halved the number of control packets. Thus, IPPv6 greatly reduced the number of control packets. No. of control packets (normalized simulation results) subnets 19 subnets 37 subnets 61 subnets Interval between sessions = 1800 s (s) Handoff interval 6. Conclusion This article described IP-based mobility management technology that enhances the mobility of mobile nodes in 4G systems. Featuring multiple interface management, active state mobility management, and dormant state mobility management, it satisfies all the key requirements, including high packet communication quality, modest control costs, and seamless mobility. Fig. 7. Reduction in number of control packets achieved by IPPv6 (simulation results). References [1] T. Ikeda, I. Okajima, and N. Umeda, Link Manager for IP-Based Mobile Communications System, IEICE, B-5-58, Sep [2] R. Kobayashi, M. Inoue, I. Okajima, and N. Umeda, Management Information Exchange which should be Commonly Shared among the Link Layers for IP-Based Mobile Communications System, IEICE, B-5-128, Sep [3] D. Johnson, C. Perkins, and J. Arkko, Mobility Support in IPv6, draft-ietf-mobileip-ipv6-24.txt, June [4] H. Soliman, C. Castelluccia, K. El Malki, and I. Bellier, Hierarchical 46 NTT Technical Review

36 Mobile IPv6 Mobility Management (HMIPv6), draft-ietf-mobileiphmipv6-08.txt, June [5] K. Omae, I. Okajima, and N. Umeda, Hierarchical Mobile IPv6 Extension for IP-based Mobile Communications System, IEICE, IN Research Group, IN , Feb [6] H. Takahashi, R. Kobayashi, I. Okajima, and N. Umeda, Transmission Quality Evaluation of Hierarchical Mobile IPv6 with Buffering Using Test Bed, Proceedings of IEEE VTC 2003 Spring, Apr [7] K. Omae, M. Inoue, I. Okajima, and N. Umeda, Hierarchical Agent Discovery for Hierarchical IP Mobility Management Protocol, IEICE, B-5-116, Mar [8] I. Okajima, N. Umeda, and Y Yamao, Architecture and Mobile IPv6 Extensions Supporting Mobile Networks In Mobile Communications, Proceedings of IEEE VTC 2001 Fall Vol. 4, Oct [9] [10] [11] ITU-T Recommendation G.729 Annex A, Reduced complexity 8kbit/s CS-ACELP speech codec, Nov [12] ITU-T Recommendation Y.1541, Network Performance Objectives for IP-Based Services, Feb [13] M. Inoue, I. Okajima, and N. Umeda, IP Paging for IP-based Mobile Communications System, IEICE, RCS2004-3, Apr Ichiro Okajima Senior Research Engineer, Wireless Laboratories, NTT DoCoMo, Inc. He joined NTT DoCoMo in He is engaged in research on 4G mobile communications systems. He is a member of the Institute of Electronics, Information and Communication Engineers (IEICE). Hideaki Takahashi Wireless Laboratories, NTT DoCoMo, Inc. He joined NTT DoCoMo in He is engaged in research on 4G mobile communications systems. He is a member of the Information Processing Society of Japan (IPSJ). Masahiro Inoue Research Engineer, Wireless Laboratories, NTT DoCoMo, Inc. He joined NTT DoCoMo in He is engaged in research on N-STAR satellite mobile communications systems and 4G communications systems. He is a member of IEICE and IEEE. Koji Omae Wireless Laboratories, NTT DoCoMo, Inc. He joined NTT DoCoMo in He is engaged in research on 4G mobile communications systems. He is a member of IEICE. Narumi Umeda Director, Communication Systems Laboratory, Wireless Laboratories, NTT DoCoMo, Inc. He received the B.E. and M.E. degrees in electronic engineering from Hokkaido University, Sapporo, Hokkaido in 1985 and 1987, respectively. He joined NTT Laboratories in In 1992, he transferred to NTT Mobile Communications Network, Inc. (now NTT DoCoMo, Inc.). He has been engaged in research on the radio link control for the personal digital cellular (PDC) system and IMT His current research interests are in the radio control for 4G mobile communications systems (beyond 3G). He was a co-recipient of the Japan Institute of Invention and Innovation (JIII) Imperial Invention Prize in 1998 and the best paper award of the International Conference on Telecommunications at ICT2002. He is a member of IEICE and IEEE. Vol. 2 No. 9 Sep