Network Mobility Protocol for Vehicular Ad Hoc Networks

Transcription

1 INTERNATIONAL JOURNAL OF COMMUNICATION SYSTEMS Int. J. Commun. Syst. 2010; 00:1 35 Published online in Wiley InterScience ( Network Mobility Protocol for Vehicular Ad Hoc Networks Yuh-Shyan Chen, Chih-Shun Hsu, Ching-Hsueh Cheng Department of Computer Science and Information Engineering, National Taipei University, New Taipei 237, Taiwan Department of Information Management, Shih Hsin University, Taipei 116, Taiwan SUMMARY The goal of the network mobility (NEMO) management is to effectively reduce the complexity of handoff procedure and keep mobile devices connecting to the Internet. When users are going to leave an old subnet and enter a new subnet, the handoff procedure is executed on the mobile device and it may break off the real time service, such as VoIP or mobile TV, due to the mobility of mobile devices. Since a vehicle is moving so fast that it may cause the handoff and packet loss problems. Both of the problems will lower down the throughput of the network. To overcome these problems, we propose a novel NEMO protocol for vehicular ad hoc networks (VANETs). In a highway, since every car is moving in a fixed direction at a high speed, a car adopting our protocol can acquire an IP address from the VANET through the vehicle-tovehicle communications. The vehicle can rely on the assistance of a front vehicle to execute the pre-handoff procedure or it may acquire a new IP address through multi-hop relays from the car on the lanes of the same or opposite direction and thus may reduce the handoff delay and maintain the connectivity to the Internet. Simulation results have shown that the proposed scheme is able to reduce both the handoff delay and packet loss rate. Copyright c 2010 John Wiley & Sons, Ltd. Received... KEY WORDS: Network mobility (NEMO), Mobile router (MR), vehicular ad hoc network, mobility management, pre-handoff, wireless network, WiMAX. Correspondence to: Department of Information Management, Shih Hsin University, Taipei 116, Taiwan, cshsu@cc.shu.edu.tw Copyright c 2010 John Wiley & Sons, Ltd. [Version: 2010/03/27 v2.00]

2 2 Y.-C. CHEN OTHER 1. INTRODUCTION Due to the advancement and popularity of the wireless technology, the dream of the ubiquitous communication has become a reality. Nowadays, users may enjoy many kinds of services through Internet anytime and anywhere. As the wireless transmission rate increases, more and more multimedia services have been provided to serve mobile users. No matter the users are in low mobility (e.g. walking or jogging) or high mobility (e.g. driving a car or taking a bus or a subway), they all can make a VoIP call, browse website, download data, watch TV, and get road traffic information or real-time weather report from Internet through wireless communications. Nowadays, most people go to work by vehicles, so vehicles become more and more important in our daily life. Since every car can be equipped with a short- or medium-range wireless transceiver, cars on the road may form a wireless network named as a vehicular ad hoc network (VANET). The VANET is a subclass of the mobile ad hoc network (MANET) that it also has no fixed topology. Vehicles may acquire information and services through the V2V (Vehicle-to-Vehicle) or I2V (Infrastructure-to-Vehicle) communications. The V2V communication is based on the Dedicated Short Range Communications (DSRC) technology; while the I2V communication is based on GPRS/3G, WiFi or WiMAX. Since the moving speed of a vehicle in the VANET is so high that it is harder to maintain a seamless handoff and a stable connectivity to the Internet. To achieve the seamless handoff for an IP-based communication, the IP of the mobile device must be assigned and reassigned efficiently. The Mobile Internet Protocol version 4 (MIPv4)[1] has been proposed by the Internet Engineering Task Force (IETF). Since MIPv4 may face problems like the shortage of IP addresses and the weak security mechanism, MIPv6 [2] is proposed by IETF to alleviate the above problems. Mobile IP has an important characteristic that it configures the IP address by neighbor discovery or autoconfiguration. There are two auto-configuration mechanisms: the stateful and stateless mechanisms. Dynamic Host Configuration Protocol (DHCP)[3], which has been adopted in both IPv4 and IPv6, is a stateful auto-configuration mechanism that each IP address is autoconfigured and managed

3 NETWORK MOBILITY PROTOCOL FOR VEHICULAR AD HOC NETWORKS 3 by a DHCP server [4][5]. The stateless auto-configure mechanism is adopted by IPv6, which creates link-local IPv6 addresses. Although MIPv6 can provide enough IP addresses and better security mechanism than MIPv4, it is not efficient enough. To improve the efficiency of MIPv6, IETF has proposed a hierarchical mobile Internet Protocol (HMIPv6)[6]. HMIPv6 adds a new component named as Mobility Anchor Point (MAP), which manages users location. MAP provides two types of location management: the macro-mobility and micro-mobility managements, the later one can improve the handoff efficiency of MIPv6. MIPv4 and MIPv6 are designed to handle the terminal mobility and are not suitable for handling the network mobility. Therefore, the Request for Comments (RFC) of IETF has extended MIPv6 to support Network Mobility (NEMO), named as the NEMO basic support protocol [7]. According to this RFC, mobile network nodes (MNNs) can only be accessed through mobile router (MR), in other words, all users are not allowed to access the base station directly. In mobile network, each MR has a home network address. When an MR moves to the communication range of a new access router, it acquires its care-of Address (CoA) from the visited network. As long as the MR acquires a CoA, it sends a binding update message to its home agent (HA). The MR provides connections to MNNs. MNNs get IP addresses through MR by DHCP and network address translation (NAT) [8]. Various wireless technologies have been provided to support a lot of services via Internet, such as IEEE and IEEE IEEE has defined the media access control layer and physical layer for wireless local area networks. It is suitable for the short- or medium-distance transmission. IEEE support two network modes, the ad hoc and the infrastructure modes. Since the coverage area of the IEEE based device is small, it may result in frequent handoff and may cause high packet loss rate and thus is not suitable for a high mobile environment. To provide a wide transmission range and a high data rate, WiMAX (Worldwide Interoperability for Microwave Access) has been proposed. WiMAX is based on the IEEE (802.16d) standard [9], that it cannot support high mobile nodes. To support high mobile nodes, IEEE e [10] has been proposed such that it provides a complete handoff procedure and a power-saving protocol.

4 4 Y.-C. CHEN OTHER In this paper, we propose a novel NEMO protocol for VANETs. Each vehicle in the VANET is equipped with an MR with two interfaces, e.g. Wi-Fi and WiMAX. The basic idea of this paper, inspired by [11][12], is shown as follows: When a vehicle is leaving the communication region of its serving base station (BS) and is moving to the boundary of the target BS s communication region, it may acquire a new IP address through multi-hop relays [13] from a car on the lanes of the same or opposite direction and it can assist a vehicle behind of it to perform the pre-handoff. A vehicle can acquire a new IP address from a vehicle on the lanes of the same direction by IP passing or it can exchange its IP with vehicles on the lanes of the opposite direction. By using the relay ability of each vehicle, the handoff efficiency can also be improved. By applying the above ideas, we can significantly reduce the handoff delay and packet loss rate. Simulations results have shown that the proposed scheme can efficiently reduce the handoff delay and packet loss rate. The remainder of this paper is organized as follows. In Section 2, related works are described. Section 3 overviews the system architecture, and the basic ideas of the proposed schemes. Section 4 describes the proposed NEMO management scheme for VANETs. Performance evaluation is presented in Section 5. Section 6 concludes this paper. 2. RELATED WORKS The handoff delay consists of layer 2 and layer 3 handoff delays. On layer 2, the mobile device spends time to scan signal strength from BS ( or ), while on layer 3, it spends time to configure IP address. Particularly, the duplicate address detection (DAD) procedure spends the most time. To achieve ubiquitous communication, devices equipped with wireless transceivers must be allowed to be roaming smoothly around the wireless networks with a seamless handoff. To achieve this, many of the works are focus on the study of reducing handoff delays and improving the efficiency of network mobility. MIPv6[2] has been proposed to support the network layer mobility. This mobile IP protocol allows a mobile node (MN) to maintain the connectivity to the Internet while moving from one subnet to another. Each MN is identified by its Home address (HoA). When connecting through a

5 NETWORK MOBILITY PROTOCOL FOR VEHICULAR AD HOC NETWORKS 5 foreign network, an MN receives the Router Advertisement, this message includes a foreign network prefix and an auto-configured Care-of address (CoA). After configuring CoA, an MN must perform the duplicate address detection (DAD) procedure to guarantee that its CoA is unique. If the CoA is usable, the MN sends its location information to its home agent (HA) to perform binding update, which intercepts packets for the MN and tunnels them to the MN s current location. In order to improve MIPv6 to support the real time handoff, Hierarchical Mobile IPv6 (HMIPv6) [6] and Fast Mobile IPv6 (FMIPv6) [14] are proposed. Two handoff modes are provided by FMIPv6, namely the predictive fast handoff mode and reactive fast handoff mode. In the predictive fast handoff mode, an MN has a previous CoA (PCoA) in the previous access router (par). When the MN has received the Router Advertisement message from the new access router (nar), it can acquire a new CoA (NCoA). When performing the binding update, the PCoA and NCoA of the MN are sent to the par. The par sends a HI (handoff initial) message to perform the DAD procedure. As long as the NCoA is available, the MN sends an HACK (handoff ACK) to the par and the par sends a fast binding ACK to the MN subsequently. When an MN enters a new subnet, it sends a fast advertisement notification (FAN) to the nar to ensure that its NCoA won t be used by the other MNs and to establish a tunnel between the par and the nar. The packets for the MN will be buffered in the nar. The cross-layer designs, including [15] and [16], aim to reduce both the packet loss rate and the handoff delay so as to achieve a seamless handoff. When a device executes the handoff procedure, it performs layer 2 handoff first and then subsequent by layer 3 handoff. When a device is moving from the serving BS to the target BS, it shall execute the handoff procedure, which may cause many signalings and thus reduces throughput. Therefore, the IETF has proposed a protocol called NEMO [7] for MIPv6 (named as MIPv6-NEMO), which provides transparency for NEMO. A survey of NEMO is conducted in[17]. Although NEMO provides a good mobility management but its handoff procedure cannot work properly in the vehicular environment. Several schemes[18] [19] [20] [21] study network mobility so as to improve the layer 3 handoff. Zhong et al.[19] propose a handoff scheme to support network mobility over e. This paper

6 6 Y.-C. CHEN OTHER considers both of the layer 2 and layer 3 handoff, but majorly for layer 3 handoff. Naoe et al. [21] propose a scheme called NEMO-SHO, which has improved the NEMO basic protocol and can be applied to the heterogeneous access network (Cellular networks and WLAN). In order to achieve a seamless vertical handoff, each mobile router (MR) is equipped with two interfaces. The first interface is for the current link and the second interface is for the new link. When the MR is moving from the serving BS to the target BS, two operations are performed. First, the MR individually establishes two bi-directional tunnels simultaneously, one is between the par and the MR via the first interface, another is between the nar and the MR via the second interface. When the new wireless link is established via the second interface, the MR sends a local binding update (LBU), which contains the PCoA, to the home agent. With the multi-interface mobile router and the multihoming concepts, a seamless handoff can be achieved. Lin et al. [20] propose a hybrid handoff scheme with multiple mobile routers called Intelligent Control Entity (ICE). It can support multihoming for NEMO and improve the handoff problem of the NEMO basic protocol. An access router can manage a number of access points (AP). The ICE can choose the best access router (AR) for an MR. To provide a seamless handoff, the ICE supports the intra-domain handoff and inter-domain handoff. A handover scheme with geographic mobility awareness (HGMA) is proposed in [32]. The historical handover patterns of mobile devices is considered in HGMA. HGMA prevents unnecessary handovers according to the received signal strength and moving speeds of mobile devices. A handover candidate selection method is proposed in HGMA for mobile devices to select a subset of WiFi access points or WiMAX relay stations for scanning. In HGMA scheme, Mobile devices prefer to stay in their original WiMAX or WiFi networks so as to prevent mobile devices from consuming too much energy on interface switching. An SIP-based mobile network architecture for supporting network mobility is proposed in [22]. This approach depends on the lower layer movement detection to trigger the high layer handoff procedure. A Session Initiation Protocol (SIP)-based cross-layer scheme is proposed in [33] so as to support seamless handover over heterogeneous networks. This scheme consists of a battery lifetime-based handover policy and a

7 NETWORK MOBILITY PROTOCOL FOR VEHICULAR AD HOC NETWORKS 7 cross-layer fast handover scheme, called the SIP-based mobile stream control transmission protocol (SmSCTP). Han et al. [18] propose a reactive handoff for NEMO. It analyzes the operations of MIPv6, including the movement detection (MD) on layer 2 and address configuration on layer 3. The reactive handoff is based on HMIPV6 and NEMO. The reactive handoff procedure optimizes the MD procedure and address configuration as follows. Initially, the MN establishes a connection to a new AP and sends the regional CoA (RCoA) to the AP. The AP sends a new-connection notification message to the AR on network layer. The AR selects an IP address from the unique address pool and acquires the local CoA (LCoA). The LCoA will be used for the MN. Finally, the AR sends a router advertisement directly to the MN. C.W. Lee et al. [23] propose a crosslayer (Layer2 and Layer 3) network mobility management and resource allocation scheme for the QoS handoff referred as HiMIP-NEMO. This approach combine the design of routing and resource allocation. The foreign mobility agent (FMA) provides the fast and reliable QoS handoff. The HiMIP-NEMO incorporates both the registration protocol and the QoS handoff protocol to reduce the latency and packet loss rate. A survey on mobility management for vehicular networks is proposed in [24]. In this paper, the requirements of mobility management for vehicular networks are identified. Existing mobility management schemes are reviewed and classified based on the vehicleto-vehicle (V2V) and vehicle-to-infrastructure (V2I) communications. This paper also discussed the differences between host-based and network-based mobility management. An architecture for interworking heterogeneous all-ip networks is proposed in [25]. This framework enables any 3G cellular technology to interwork with a broadband wireless access system or a wireless local area network. Several performance metrics such as vertical handoff delay, transient packet loss, jitter and signaling cost relating to vertical handoff management are investigated in this paper. The problem of IP mobility and its specific requirements in vehicular networks are addressed in [26]. This paper provides a qualitative comparison among the existing IP mobility solutions. Their improvements and weaknesses with respect to the current standard are also described. Recently, many researches about proxy mobile IPv6 (PMIPv6) have been proposed. Bernardos et al. [27] describe PMIPv6 and its current trends in standardization. Three extensions for

8 8 Y.-C. CHEN OTHER PMIPv6, namely, flow mobility, multicast and network mobility support, are considered by the standardization bodies to enhance the basic protocol. Recently, the draft of the PMIPv6- based solution for distributed mobility management is proposed in [28]. This solution brings the mobility anchor closer to the MN. The Mobility Anchor and Access Routers (MAARs) provide IP connectivity to a set of MNs and are also the mobility managers for those MNs. An inter- LMA handoff and location management scheme using core-edge separation network for global mobility management is proposed in [29]. A PMIPv6-based NEMO (P-NEMO) is presented in [30] to maintain the Internet connectivity of vehicles and it does not participate in the management of location update while vehicles are moving. To improve handover performance, an extension protocol of P-NEMO named as fast P-NEMO (FP-NEMO) is developed. FP-NEMO anticipate the handovers of vehicle by utilizing wireless L2 events. The handover of the vehicle is prepared before the vehicle attaches to the new access network by the mobility service provisioning entities. Qualitative and quantitative analyses of the host-based (i.e. MIPv6) and network-based mobility management (i.e. PMIPv6) approaches are presented in [31]. The authors investigate a comparison among the existing mobility support protocols. Most of the above works try to reduce the layer 2 or layer 3 handoff delay based on NEMO, so as to provide a seamless handoff. As VANETs become more and more popular, NEMO management becomes more and more important. Since vehicles are moving so fast, we must improve the NEMO basic support protocol for VANETs. Hence, we propose a novel network mobility management scheme for VANETs so as to reduce the macro mobility handoff time and packet loss rate. 3. PRELIMINARY In this section, we first describe the assumptions and system architecture, and then we explain the challenges and the basic idea of this work.

9 NETWORK MOBILITY PROTOCOL FOR VEHICULAR AD HOC NETWORKS 9 HA nar Internet par CN WLAN Access Point WLAN Access Point CN: corresponding node HA: home agent par: previous access router nar: new access router MR: mobile router connection Figure 1. System architecture 3.1. System Architecture A novel network mobility protocol for VANETs is proposed in this paper. Network mobility protocol for VAVET assumes that each of the vehicle is equipped with at least a mobile router. The system architecture of our protocol is shown in Fig.1. The scenario of our protocol is a highway with four lanes. Our protocol is based on the network mobility (NEMO) concept and combine the property of multi-hop VANETs. In WiMAX or Wi-Fi environment, the base station (BS) (or Access point (AP)) is managed by the access router (AR). The home agent (HA) records the vehicle s new location and the Correspondence Node (CN) serves as a remote server e.g. FTP, Web server etc. The car and bus can connect to the Internet through the mobile router with the wireless communication technology e.g. IEEE or In our system architecture, the mobile router has a special function to assist the pre-handoff procedure. When a bus wants to perform the handoff procedure via the cooperative vehicle s mobile router(cv-mr), which can detect neighboring BS s advertisement message. Since the bus is still not in the coverage area of the target BS, it must find the other vehicle s mobile router, which is in the overlay area of the serving BS s coverage area and target

10 10 Y.-C. CHEN OTHER Layer 3 mobility Layer 3 mobility MAC MAC MAC MAC MAC MAC MAC PHY MR NIC PHY Wi-Fi PHY MN NIC PHY MR NIC PHY Wi-Fi PHY MN NIC Wi-Fi PHY MN NIC MR NIC: Mobile Router Network Interface Card MN NIC: Mobile Node Network Interface Card Figure 2. Mobile router protocol stacks. BS s coverage area, to assist the handoff. As soon as the bus enters the target BS s coverage area, the mobile router can perform partial layer 3 handoff procedure. The mobile router has two interfaces, including an outgoing interface via WiMAX and an incoming interface via Wi-Fi. The communication between vehicles is via the Wi-Fi interface. Fig.2 shows the protocol stack of the mobile router. The protocol stacks of the vehicle and bus are shown as follows: 1) The protocol stack of a vehicle: Each vehicle is equipped with a mobile router, which is equipped with two communication interfaces, including the WiMAX and Wi-Fi interfaces. The mobile network node is equipped with only one communication interface, the Wi-Fi interface. Since the backbone is the WiMAX architecture and the mobile network node is not equipped with the WiMAX interface, the mobile network node connect to the MR via the Wi-Fi interface and then connects to the Internet via the MR s WiMAX interface. 2) The protocol stack of a bus: Each bus is equipped with one or more mobile routers, which are equipped with two communication interfaces, including the WiMAX and Wi-Fi interfaces. The mobile network node decides to adopt one hop or multi-hop communications to the

11 NETWORK MOBILITY PROTOCOL FOR VEHICULAR AD HOC NETWORKS 11 FMR RMR PDA PDA (a) (b) PDA (c) PDA (d) Figure 3. Network mobility scenario: (a) NEMO on a real Bus (b)nemo on a Virtual Bus with two vehicles (c) NEMO on a Virtual Bus with three vehicles (d) NEMO on a Virtual Bus with more than three vehicles mobile router according to the distance between the mobile network node and the mobile router. If the mobile network node is far from the mobile router, it communicates with the mobile router via multi-hop communications. Otherwise, it may adopt the one hop communication to directly communicate with the mobile router. The network mobility scenarios of the proposed protocol are shown in Fig.3. As shown in Fig.3(a), the real bus is equipped with two mobile routers, the front mobile router (FMR) and the rear mobile router (RMR). The FMR and RMR are connected by wires. The users on the bus connect to the RMR either directly or via multi-hop wireless communications. As shown in Fig.3(b), a front vehicle and a following small bus may form a virtual bus. The MR of the front vehicle serves as an FMR, while the MR of the following small bus serves as an RMR. However on a virtual bus, the FMR and RMR are connected via wireless communication. Fig. 3(c) and (d) shows two larger virtual buses. Each of the virtual bus consists of more than two vehicles. The MRs of the vehicles on the virtual bus are also connected via wireless communication.

12 12 Y.-C. CHEN OTHER On a highway, since the vehicles are moving so fast that it may pass several base stations within a few seconds. If the handoff procedure cannot finish in time, disconnection may occur. To solve this problem, we adopt the bus-like model as mentioned in the previous paragraph. The FMR uses the relay function, which has been defined in IEEE j, to assist the RMR to perform the prehandoff procedure and thus reduces the handoff delay of the RMR Basic Idea and Challenges When a mobile node moves to a new subnet, the mobile router (MR) will receive a broadcast packet from the target BS and most importantly perform the handoff procedure. The traditional handoff procedure includes two parts, the layer 2 and layer 3 handoff procedures. The handoff procedure contains signal measurement, network layer movement detection, DAD procedure and registration. The DAD procedure is time consuming and thus will cause the link to be disconnected. Although NEMO does not require every mobile node to perform the layer 2 and layer 3 handoff, but it can still perform the above works through an MR. In VANETs, since the vehicle is moving so fast that the handoff procedure must be simplified. Our protocol, based on IP passing [12], can acquire IP address faster than the traditional handoff procedure. The basic idea of our protocol is to perform the fast handoff by the assistance of other vehicles on the VANET. Our idea comes from a real bus mentioned in Fig. 3(a). The bus is assumed to be equipped with the Wi-Fi and WiMAX interfaces, where the Wi-Fi interface is for V2V communication and the WiMAX interface is used to connect to the Internet. If there is any vehicle on the lanes of the opposite direction, the mobile router may acquire its IP address from a vehicle on the lanes of the opposite direction. For example in Fig.4, the bus has two MRs, called BMR 1 and BMR 2. BMR 1 and BMR 2 are connected by a wire and thus causes a few end to end delay. When BMR 1 receives the advertisement of the target BS, it informs BMR 2 to prepare the handoff procedure. BMR 2 sends a request message called the get IP address packet to ask BMR 1 to assist its pre-handoff, and then BMR 1 replies an allowance message to BMR 2. As shown in Fig.4(a), when a bus is entering a new subnet, BMR 1 will assist BMR 2 to get its IP address. Since there is no any vehicle on the lanes

13 NETWORK MOBILITY PROTOCOL FOR VEHICULAR AD HOC NETWORKS 13 HA nar Internet par CN WLAN Access Point WLAN Access Point BMR2 BMR1 PDA BMR1 BMR2 PDA (a) HA nar Internet par CN WLAN Access Point WLAN Access Point CMR1 CMR2 BMR1 BMR2 PDA PDA (b) Figure 4. (a) Acquire IP address from the lanes of the opposite direction. (b)acquire IP address from the lanes of the same direction. of the same direction released its IP address, BMR 1 may acquire the IP address from BMR 1 or BMR 2, which are the mobile routers of the bus on the lanes of the opposite direction. Initially, all mobile network nodes (PDA, ) receive data form the correspondence node (CN) via the home agent (HA) and BMR 2. If BMR 2 performs pre-handoff, it will send a pre-handoff request to BMR 1. When two buses on the opposite direction are entering the new subnet. Since they cannot acquire IP addresses from the vehicle on the lanes of the same direction, the MRs of the two buses BMR 1 and BMR 1 exchange their IP addresses.

14 14 Y.-C. CHEN OTHER However, if there is no any vehicle on the lanes of the opposite direction, the mobile router may acquire the IP address from the vehicle on the lanes of the same direction. When the mobile router of a vehicle (CMR 1 ) is going to leave the target BS s communication region, it passes its IP address to CMR 2, which is the MR of the back vehicle, by multi-hop communications. For example in Fig.4(b), when CMR 1 is going to leave the target BS s communication region, CMR 1 passes its IP address to BMR 1 through CMR 2 by multi-hop communications (CMR 1 CMR 2 BMR 1 ). The concept of our protocol is similar to [15], [11] that the cooperative node can assist other mobile nodes to perform pre-handoff. However, the real bus scheme has two drawbacks: First, the distance between BMR 1 and BMR 2 is too short to reduce enough handoff latency through the prehandoff procedure performed by BMR 1 and thus cannot provide a seamless handoff when the speed of the vehicle is high. Second, each vehicle needs to equip with two MRs in the real bus scheme and thus increases the hardware cost. To improve the real bus scheme, we propose a virtual bus scheme, which consists of at least two vehicles. Each vehicle only needs to equip with an MR and thus reduces the hardware cost. Since the distance between the front vehicle and rear vehicle is longer than the distance between BMR 1 and BMR 2, the pre-handoff procedure performed by the front vehicle can be done long before the rear vehicle leaving the serving BS and thus reduces more handoff latency. 4. A NETWORK MOBILITY PROTOCOL FOR VANETS Our NEMO protocol contains two algorithms, one for the real bus and one for the virtual bus. Every algorithm is split into the following phases: the information collecting phase, the fast IP acquiring phase, the cooperation of mobile routers phase, the make before break phase, and the route redirection phase. In the information collecting phase, each vehicle uses its mobile router to broadcast its own and its neighboring vehicles locations, moving speeds, and directions periodically. Besides, it should also rebroadcast the messages it received according to the TTL (time to live) of the messages. If the TTL of a message is greater than 0, then the message should be rebroadcast. The TTL is set according to

15 NETWORK MOBILITY PROTOCOL FOR VEHICULAR AD HOC NETWORKS 15 the intended size of the virtual bus and the communication range of the base station. After collecting the information of the nearby vehicles, a vehicle can realize the neighboring vehicles locations, moving speeds, and directions and thus can group proper vehicles to form a virtual bus and can choose a proper cooperative vehicles to assist it to perform the prehandoff procedure at the proper moment Algorithm A: The NEMO Scheme for a Real Bus In this algorithm, the bus is equipped with two mobile routers, the first mobile router (BMR 1 ) performs the pre-handoff procedure and the second mobile router (BMR 2 ) serves the mobile nodes and maintains the connectivity to the Internet. If there is at least a vehicle on the lanes of the opposite direction that has an available IP address, the per-handoff procedure is performed as follows: S1: The mobile network nodes connect to the Internet via BMR 2. S2: BMR 2 ACQ IP REQ = BMR 1 : If there is at least a vehicle on the lanes of the opposite direction, the mobile router acquires its IP address from a vehicle on the lanes of the opposite direction. When a bus enters a new subnet, BMR 1 assists BMR 2 to get an IP address. BMR 1 acquires the IP address from a bus on the lanes of the opposite direction. For example in Fig.5(a), all mobile network nodes (PDA, ) receive data form CN via the home agent (HA) and BMR 2. If BMR 2 is going to perform the pre-handoff procedure, BMR 2 sends the pre-handoff request to BMR 1. Similarly, when a bus on the lanes of the opposite direction enters a new subnet, BMR 2 is going to perform the pre-handoff procedure and it sends an acquire IP address request to BMR 1. If there is no any available IP address on the lanes of the same direction. Under such circumstance, two MRs (BMR 1 and BMR 2 ) of the two buses exchange their IP addresses. When there is no any vehicle on the lanes of the opposite direction, but there is at least a vehicle on the lanes of the same direction ahead of the bus, the per-handoff procedure is performed as follows: S1: The mobile network nodes connect to the Internet via BMR 2.

16 16 Y.-C. CHEN OTHER HA nar Internet par CN WLAN Access Point WLAN Access Point BMR2 BMR1 PDA BMR1 BMR2 PDA (a) HA nar Internet par CN WLAN Access Point WLAN Access Point CMR1 CMR2 BMR1 BMR2 PDA PDA (b) Figure 5. (a)acquire IP address from the vehicle on the lanes of the opposition direction. (b)acquire IP address from the vehicle on the lanes of the same direction. S2: When there is no any vehicle on the lanes of the opposite direction, the mobile router acquires the IP address from a vehicle on the lanes of the same direction. When BMR 1 has received the advertisement of the target BS, it informs BMR 2 to prepare for the handoff procedure. The BMR 2 sends a request message called the get IP address packet to BMR 1 and asks BMR 2 to assist the pre-handoff procedure. BMR 1 then reply an allowance message to BMR 2. When the mobile router of a vehicle (CMR 1 ) is going to leave the target BS s communication

17 NETWORK MOBILITY PROTOCOL FOR VEHICULAR AD HOC NETWORKS 17 region, it passes its IP address to the MR of another vehicle (CMR 2 ) behind of it by multihop communications. For example in Fig.5(b), all mobile network nodes (PDA, ) receive data form CN via the home agent (HA) and BMR 2. When BMR 2 has received the ready handoff message from BMR 1, BMR 2 performs the handoff procedure by sending the acquire IP address request to BMR 1. The IP address of CMR 1 is passed to BMR 1 by multi-hop communications (CMR1 CMR 2 BMR 1 ). If the IP address cannot be acquired through the above procedures, it acquires its IP address through the DHCP procedure. After acquired an IP address, the mobile router performs the following steps. S3: BMR 1 PRE BU = HA : After BMR 1 got an IP address, BMR 1 performs pre-binding update to the home agent (HA) when BMR 2 is added into the new subnet, as show in Fig. 6(a). S4: BMR 2 Disconnect = par : When BMR 1 helps BMR 2 to perform the pre-handoff procedure, BMR 2 establishes a connection to HA. Even BMR 2 has entered the target BS s communication region, the old connection between the serving BS and BMR 2 still exists until BMR 2 is away from the serving BS s communication region, as show in Fig.6(b). S5: BMR 2 RR = CN : The network mobility route redirection is similar to MIPv6. When BMR 2 has moved to the communication region of the target BS, all of the packets from CN to HA and HA to nar to mobile network nodes are via BMR 2, which may cause the network to be overloaded. Therefore, NEMO route redirection must be performed with CN. The packets form CN is directly transferred to BMR 2, does not need to via HA again, as shown in Fig.7. The message flow diagram of the real bus scheme is shown in Fig. 8, which summarizes the steps of algorithm A Algorithm B: Virtual Bus network mobility protocol In VANETs, how to acquire IP addresses immediately is a very important issue. In this subsection, the bus is only equipped with a mobile router but still with WiMAX and Wi-Fi interfaces. Two or more neighboring vehicles on the lanes of the same or opposite direction may be grouping as

18 18 Y.-C. CHEN OTHER HA nar Internet par CN WLAN Access Point WLAN Access Point BMR1 BMR2 PDA PDA (a) HA Internet CN nar par WLAN Access Point WLAN Access Point BMR1 BMR2 PDA PDA (b) Figure 6. (a) Binding update by BMR 2. (b) Disconnect the old link. HA Internet CN nar par WLAN Access Point WLAN Access Point BMR1 BMR2 PDA PDA Figure 7. NEMO route redirection.

19 NETWORK MOBILITY PROTOCOL FOR VEHICULAR AD HOC NETWORKS 19 Serving-BS RMR2 FMR1 Target-BS Release IP of MR HA CN par nar Packet forwarding Packet forwarding Link layer handoff procedure Request Reply Acquire unique IP Request Reply(Fail) Acquire IP Binding Update Binding ACK Packet tunnel Disconnect Packet forwarding Binding Update Binding Update ACK Packet forwarding Figure 8. The message flow diagram of the real bus scheme. a virtual bus. If no vehicle on the lanes of the same direction is at the target BS s communication region and the mobile router of the virtual bus cannot acquire the target BS s IP address at the serving BS communication region. In the worst case, a vehicle can get its IP address from the DHCP server but this approach will spend too much time and cause a longer handoff delay. Without loss of generality, assume that in the serving BS s communication region, three neighboring vehicles on the lanes of the same direction are grouping as a virtual bus. Assume that there are three vehicles (CMR 1, CMR 2 and CMR 3 ) on the lanes of the opposite direction, the vehicle may perform the pre-handoff procedure as follows. S1: The mobile network nodes connect to the Internet via virtual BMR 3. S2: Virtual BMR 3 ACQ IP REQ = Virtual BMR 1 : The mobile router (BMR 3 ) of the virtual bus sends an acquire IP address request to the front mobile router (BMR 2 ) of the virtual bus. Since BMR 2 is still not at the target BS s communication region, it forwards the request to BMR 1. When BMR 1 receives the request, it accepts the request message as shown in Fig.9(a). S3: CMR 3 ACQ IP REQ = CMR 1 : On the lanes of the opposite direction, there are three vehicles moving from the serving BS to the target BS s communication region, the mobile router

20 20 Y.-C. CHEN OTHER (CMR 3 ) sends a request message to the front mobile router (CMR 2 ) of the front vehicle. CMR 2 is still at the srving BS s communication region, it forwards the request to CMR 1. At this moment, CMR 1 accepts the request message because it is at the communication regions of both the serving BS and the target BS. When CMR 1 and BMR 1 are both at the same communication regions of both the serving BS and the target BS, CMR 2 and BMR 2 can exchange their IP addresses via CMR 1 and BMR 1 and thus can previously acquire their IP addresses, as shown in Fig.9(a). When there is no any vehicle on the lanes of the opposite direction but there are some front vehicles on the lanes of the same direction, the pre-handoff procedure of the first two steps are the same as the previous case. The third step is performed as follows: S3: CMR 1 IP PASSING = Virtual BMR 1 : If the mobile router of a vehicle (CMR 1 ) is going to leave the target BS s communication region, it releases its IP address and passes the IP address back to the virtual BMR 1 via multi-hop communications (CMR 1 CMR 2 BMR 1 ) or it may release its IP address back to the DHCP server as shown in Fig.9(b). If the IP address cannot be acquired through the above procedures, it acquires the IP address through the DHCP procedure. After acquired the IP address, the mobile router performs the following steps. S4: Virtual BMR 1 BU = HA: BMR 3 has finished acquiring a unique IP address. When BMR 3 enters the new BS s communication region, BMR 3 sends a binding update to HA. HA receives the message and responds a binding ACK as shown in Fig 10. S5: Virtual BMR 3 Bitunnel = HA: When the mobile router (BMR 3 ) is moving to the new BS s communication region but still not leaving the communication region of the serving BS, BMR 3 maintains the old link to the serving BS and establishes a new link to the new BS so as to achieve soft-handoff. BMR 3 maintains the old bidirectional tunnel and establishes a new bidirectional tunnel to the new BS. The packet will be delivered from CN to virtual BMR 3 via HA.

21 NETWORK MOBILITY PROTOCOL FOR VEHICULAR AD HOC NETWORKS 21 HA nar Internet par CN WLAN Access Pointer WLAN Access Pointer CMR3 CMR2 CMR1 BMR1 BMR2 BMR3 PDA (a) HA nar Internet par CN WLAN Access Pointer WLAN Access Pointer CMR1 CMR1 BMR1 BMR2 BMR3 PDA (b) Figure 9. (a) A vehicle of the virtual bus acquire IP address from the vehicle on the lanes of the opposite direction. (b)a vehicle of the virtual bus acquire IP address from the vehicle on the lanes of the same direction. S6: Virtual BMR 2 Disconnect = HA: When the mobile router BMR 3 of the virtual bus detects that the target BS s signal strength has reached the handoff threshold, BMR 3 disconnects the link to the old BS, as shown in Fig 11(a). S7: Virtual BMR 2 BU = CN: The mobile router (BMR 3 ) of the virtual bus sends the binding update to CN. It informs CN with CoA and mobile network prefix. This can redirect the route between the mobile router and CN as shown in Fig.11(b).

22 22 Y.-C. CHEN OTHER HA nar Internet par CN WLAN Access Pointer WLAN Access Pointer BMR1 BMR2 BMR3 PDA Figure 10. NEMO Binding update. HA Internet CN nar par WLAN Access Pointer WLAN Access Pointer BMR1 BMR2 BMR3 PDA (a) HA Internet CN nar par WLAN Access Pointer WLAN Access Pointer BMR1 BMR2 BMR3 PDA (b) Figure 11. (a) NEMO Make Before Break.(b) Route redirection Phase.

23 NETWORK MOBILITY PROTOCOL FOR VEHICULAR AD HOC NETWORKS 23 BMR Virtual Bus CMR CMR Serving-BS Target-BS Releases IP of MR HA CN PDA par nar Packet forwarding Link layer handoff procedure Request Reply Request Reply Acquire unique IP Request Reply (Fail) Request Reply (Fail) Acquire IP Binding Update Binding ACK Packet tunnel Packet forwarding Disconnect Binding Update Binding Update ACK Packet forwarding Figure 12. The message flow diagram of the virtual bus scheme. The message flow diagram of the virtual bus scheme is shown in Fig. 12, which summarizes the steps of algorithm B. According to the above description, we can see that the proposed virtual bus scheme can reduce the handoff latency and packet loss rate by the pre-handoff procedure performed by the front vehicle in the virtual bus and thus the handoff can be done long before the rear vehicle leaving the serving BS. Besides, the proposed virtual bus scheme can also reduce the hardware cost such that each vehicle is only equipped with an MR. 5. SIMULATION RESULTS To evaluate the performance of our protocol, we simulate the original network mobility basic support protocol [7], the fast network mobility protocol [14] and our protocol on Network Simulator-2 (ns- 2)[34]. For the simplicity of describing the simulation results, the original network mobility basic support protocol is denoted as NEMO, the fast network mobility protocol is denoted as FNEMO, our protocol based on a real bus is denoted as Bus, and our protocol based on a virtual bus is denoted as

24 24 Y.-C. CHEN OTHER Virtual Bus. Each simulation result is derived from the average of 10 simulations. The performance metrics that will be analyzed in our simulations are shown as follows: Handoff latency The handoff latency is defined as the interval that the last packet is received from the par to the time that the first packet is received from the nar. Packet loss rate The total number of lost packets over the total number of transmitting packets. Handoff jitter The handoff jitter is count during the handoff time. Assume that there are three consecutive packets, denoted as P i 2, P i 1 and P i, received by the mobile router. Let T i 2, T i 1 and T i denoted as the time to receive packets P i 2, P i 1 and P i, respectively. The handoff jitter is defined as HJ j 2 = (T i - T i 1 )-(T i 1 -T i 2 )=T i -2T i 1 +T i 2. Message overhead The total number of IP-passing packets and the packets to discover CV-MR (cooperative vehicle mobile router). Throughput The throughput is defined as the total amount of mesages received by the destination per second. Simulation parameters are shown in Table I Handoff latency Fig. 13 shows the handoff latencies of the NEMO, FNEMO, Bus, and Virtual Bus schemes. The handoff latency of the Virtual Bus scheme is the lowest, followed by the Bus, FNEMO, and NEMO schemes. Fig. 13(a) illustrates the impact of vehicle speed to handoff latency. In this simulation, there are total 50 vehicles. As the vehicle speed increases, the handoff latency also increases. Both of the Bus and Virtual Bus schemes perform better than the FNEMO and NEMO schemes because the mobile router in our scheme can perform pre-handoff procedure and thus reduces the handoff

25 NETWORK MOBILITY PROTOCOL FOR VEHICULAR AD HOC NETWORKS 25 Table I. Simulation parameters Network size Number of vehicles Vehicle speed Transmission range of WiMAX Transmission range of WLAN Packet size Packet rate Simulation Time 1000m 1000m km/h 1000m 300m 320 bytes 100 packets/sec 200sec latency. The handoff latency of the Virtual Bus scheme is lower than that of the Bus scheme because the MR on the front vehicle of the virtual bus can perform the pre-handoff procedure earlier than the FMR on the real bus and thus reduces more handoff latency. The FNEMO performs better than the NEMO scheme because the predictive fast handoff mode and reactive fast handoff mode of FNEMO can reduce a few handoff latency of the NEMO scheme. Fig. 13(b) illustrates the impact of vehicle density to handoff latency. The number of vehicles is tuned between Each vehicle is moving at a fixed speed (50km/h). As the vehicle density increase, the handoff latency also increases because higher vehicle density may incur more congestions and contentions and thus increases the handoff latency. The Virtual Bus scheme still performs the best, followed by the Bus, FNEMO, and NEMO schemes. Fig. 13(c) illustrates the impact of hop count of IP passing to handoff latency. As the hop count increases, the handoff latency also increases because more hop count may bring longer time to pass IP to the target MR and thus increases the handoff latency. The Virtual Bus scheme still performs better than the Bus scheme. Fig. 13(d) illustrates the impact of the length of virtual bus (hops) to handoff latency. As the length of virtual bus (hops) increases, the handoff latency also increases because longer length of the virtual bus may bring longer time to pass IP to the target MR. Three vehicle speeds are considered, including 20km/h, 60km/h, and 100km/h. The lower the vehicle speed is the lower the handoff latency becomes.

26 26 Y.-C. CHEN OTHER Figure 13. (a) Handoff latency vs. vehicle speed. (b) Handoff latency vs. vehicle density.(c) Handoff latency vs. hop count of IP passing (hops). (d) Handoff latency vs. Length of virtual BUS (hops) 5.2. Packet loss Rate Fig. 14 shows the packet loss rates of the NEMO, FNEMO, Bus, and Virtual Bus schemes. The packet loss rate of the Virtual Bus scheme is the lowest, followed by the Bus, FNEMO, and NEMO schemes. As mentioned in the previous subsection, the handoff latency of the Virtual Bus scheme is the lowest. Therefore, the vehicle adopting the Virtual Bus scheme, whose connection to the Internet is less likely to be broken during the handoff procedure, incurs lower packet loss rate. Fig. 14(a) illustrates the impact of vehicle speed to packet loss rate. As the vehicle speed increases, the packet loss rate also increases because higher vehicle speed may incur more frequent handoff

27 NETWORK MOBILITY PROTOCOL FOR VEHICULAR AD HOC NETWORKS 27 and can tolerate lower handoff latency and thus causes the connection to the Internet more easily to be broken during the handoff procedure. Hence, on a high speed vehicle, the packet is more likely to be lost during the handoff procedure. Both of the Bus and Virtual Bus schemes perform better than the FNEMO and NEMO schemes because our schemes incur less handoff latency. The packet loss rates of the Bus scheme grows faster than that of the Virtual Bus scheme because higher vehicle speed can only tolerate low handoff latency and the handoff latency of the Bus scheme is higher than that of the Virtual Bus scheme and thus incurs higher packet loss rate. Fig. 14(b) illustrates the impact of vehicle density to packet loss rate. As the vehicle density increases, the packet loss rate also increases because higher vehicle density may incur higher handoff latency and thus the connection to the Internet is more likely to be broken. The Virtual Bus scheme still performs the best, followed by the Bus, FNEMO, and NEMO schemes. Fig. 14(c) illustrates the impact of hop count of IP passing to packet loss rate. As the hop count increases, the packet loss rate also increases because more hop count may incur higher handoff latency and thus causes more lost packets. The Virtual Bus scheme still performs better than the Bus scheme. Fig. 14(d) illustrates the impact of the length of virtual bus (hops) to packet loss rate. As the length of virtual bus (hops) increases, the packet loss rate also increases because a longer virtual bus may incur higher handoff latency. Three vehicle speeds are considered, including 20km/h, 60km/h, and 100km/h. The lower the vehicle speed is the lower the packet loss rate becomes Handoff jitter Fig. 15 shows the handoff jitters of the NEMO, FNEMO, Bus, and Virtual Bus schemes. Similar to the results of previous subsections, the handoff jitter of the Virtual Bus scheme is the lowest, followed by the Bus, FNEMO, and NEMO schemes. Fig. 15(a) illustrates the impact of vehicle speed to handoff jitter. As the vehicle speed increases, the handoff jitter also increases. The Virtual Bus scheme performs the best, followed by the Bus, FNEMO, and NEMO schemes. Fig. 15(b) illustrates the impact of vehicle density to handoff jitter. As the vehicle density increase, the handoff jitter also increases. The Virtual Bus scheme still performs the best, followed by the Bus, FNEMO,

28 28 Y.-C. CHEN OTHER Figure 14. (a)packet loss rate vs. vehicle speed. (b) Packet loss rate vs. vehicle density.(c)packet loss rate vs. hop count of IP passing (hops). (d) Packet loss rate vs. Length of virtual BUS (hops) and NEMO schemes. Fig. 15(c) illustrates the impact of hop count of IP passing to handoff jitter. As the hop count increases, the handoff jitter also increases. The Virtual Bus scheme still performs better than the Bus scheme. Fig. 15(d) illustrates the impact of the length of virtual bus (hops) to handoff jitter. As the length of virtual bus (hops) increases, the handoff jitter also increases. Three vehicle speeds are considered, including 20km/h, 60km/h, and 100km/h. The lower the vehicle speed is the lower the handoff jitter becomes.

29 NETWORK MOBILITY PROTOCOL FOR VEHICULAR AD HOC NETWORKS 29 Figure 15. (a)handoff jitter vs. vehicle speed. (b)handoff jitter vs. vehicle density.(c)handoff jitter vs. hop count of IP passing (hops). (d) Handoff jitter vs. Length of virtual BUS (hops) Message overhead Since the NEMO and FNEMO schemes have no IP-passing packets and do not need to discover CV-MR (cooperative vehicle mobile router). Their message overhead is zero. Therefore we do not compare the proposed schemes with the NEMO and FNEMO schemes in term of the message overhead. Fig. 16 shows the message overhead of the Bus and Virtual Bus schemes. The message overhead of the Bus scheme is lower than that of the Virtual Bus scheme because on a real bus, the transmission between the RMR and FMR is single hop and via wires; while on a virtual bus, the transmission between the RMR and FMR may be more than a hop and it is wireless. Fig. 16(a) illustrates the impact of vehicle speed to message overhead. As the vehicle speed increases, the message overhead of the Virtual Bus scheme also increases because of the nature of multi-hop