Enhanced Lightweight Medium Access (el-mac) Protocol for Wireless Sensor Network

Enhanced Lightweight Medium Access (el-mac) Protocol for Wireless Sensor Network 1 Rozeha A. Rashid, 1 L.A.Latiff, 1 W.M.A Wan Embong, 1 N. Fisal, 1 S.H.S. Ariffin, 1 S.K. S. Yusof, 2 Anthony Lo 1 Telematic Research Group, Faculty of Electrical Engineering, Universiti Teknologi Malaysia 2 Wireless & Mobile Communications Group, Delft University of Technology, Netherlands {rozeha sheila sharifah kamilah}@fke.utm.my, liza@ic.utm.my, ariffehsan@gmail.com, A.C.C.Lo@tudelft.nl Abstract Wireless sensor network (WSN) is increasingly being used in a variety of applications which include habitat monitoring, smart health care system, building automation, to name a few. Many approaches were developed for all protocol layers, but an energy-efficient Medium Access Control (MAC) layer remains a key design challenge. MAC with scheduled based architecture provides greater advantage over other designs, such as contentionbased and frequency division multiple access (FDMA), in terms of minimizing packet collision, overhearing, idle listening, and over emitting. Reliable and energy efficient data transmission are required to prolong the network lifetime. This paper presents the testbed development of an enhanced lightweight medium access (el-mac) protocol which introduces distributed time slot assignment and slotting communication mechanism. Therefore, with el-mac, idle listening, overhearing and hidden terminal will be eliminated where nodes transmit in its own time slot and sleep in other time slot if there is no activity. This will reduce energy consumption as nodes are active when transmitting and receiving and idle only in the beacon session. The testbed developed consists of 9 TelosB sensor nodes programmed with TinyOS. NesC programming language was used to implement the protocols in the WSN module. Experimental results were compared to the results obtained from simulation. As expected, there is a slight degradation in throughput and packet received ratio in the experiment but is consistent for all values. This concludes that the developed testbed reflects the el- MAC protocol and has been successfully implemented. Keywords-component; Medium Access Layer (MAC), Time Division Multiple Access (TDMA), Contention-based MAC, Wireless Sensor Network (WSN), Distributed time slot assignment. I. INTRODUCTION Wireless sensor network (WSN) is composed of a collection of sensor nodes that interact with each other intentionally to gather information from the surveillance area. Sensor nodes such as TelosB, MicaZ and Sensinode are normally assembled from a low cost microcontroller, low power radio transceiver, data logger and a number of sensors. They are designed to support unattended operation for long duration, usually in remote areas, in smart building or even in hostile environments. WSN applications such as environmental monitoring [1], object tracking [2] and intelligent buildings [3] require a reliable data transmission and can endure long periods of operation. However, there are limited features in sensor node architecture which include low processing capabilities and low memory capacities. These constraints have motivated extensive research to design mechanisms that can lengthen the sensor nodes operating time and relatively prolongs the network lifetime. Medium access control (MAC) layer is a part of Open System Interconnection (OSI) model [4]. This layer is responsible to managing the medium accessibility in order to minimize collision among transmitting packets. Packet collision requires node to retransmit the packet, hence consuming additional energy. Since this layer controls the physical (radio transceiver) layer, it has a large impact on the overall energy consumption and hence, the lifetime of a node. There are several location-dependent carrier sensing issues that need to be addressed in designing MAC protocol. First is the hidden nodes problem where the nodes falsely assumed that the channel is in idle condition and start the transmission. Hence, it will result in high probability of data collision. While the exposed nodes problem occur when nodes are in transmitter range but out of receiver range. Lastly is the capture effect problem where nodes can receive one of two simultaneous transmissions. The hidden nodes problem causes retransmission of the collided data. Therefore, more energy is required, thus, inefficiency of energy usage. While the exposed nodes problem results in packet 296

overhearing, where node is forced to receive data that is not destined to it. The biggest source of energy consumption is idle listening. It occurs when a node is required to continuously listen to the channel in order to receive a potential packet from its neighboring nodes. Energy inefficiency caused by the idle-listening problem and high collision probability can be avoided in Time Division Multiple Access (TDMA) based MAC protocols. In this paper, a novel lightweight MAC protocol with distributed TDMA mechanism, named el- MAC, is presented. It possesses an active/sleep mechanism for efficient energy usage with predefined duty cycle. In addition, a distributed time slot scheduling algorithm called Adaptive Multi-timeslot Allocation (AMTA) is developed to reduce transmission latency and to adapt to traffic condition by allowing a node to control multiple timeslot per frame. The rest of the paper is organized as follows. Section II describes several related works on MAC layer protocols. Our enhanced lightweight MAC protocol design is presented in Section III. Section IV describes the development of the el-mac test-bed in terms of hardware and software platforms. The performance of test-bed results are compared with simulations in Section V. Finally, conclusions are drawn in the last section. II. RELATED WORKS In general, MAC protocol can be categorized into two types; random access and conflict-free multiple access. Traditional MAC protocols such as ALOHA [5], CSMA [6], and MACA [7], are designed based on contentionbased random access approach. The classic ALOHA protocol uses simple transmission mechanism where node transmits a packet when it is generated. However, its simplicity comes at an expense of very high probability of packet collision; hence increases the energy expenditure due to packet retransmission. Therefore, Carrier Sense Multiple Access (CSMA) protocol is developed [6] with the objective of minimizing collision by implementing a small time for channel listening in order to detect channel activity. However, the protocol cannot solve the hidden terminal problem which normally occurs in ad-hoc networks where the radio range is not large enough to allow communication between arbitrary nodes and two or more nodes may share a common neighbor while being out of each other s reach. The MACA protocol introduces a three-way handshake mechanism to make hidden nodes aware of upcoming transmission, so collision at neighboring nodes can be avoided. However, the handshaking mechanism causes overhead on control packet. All the mentioned protocols above require all nodes to continuously listen to the channel due to unpredictable packet transmission by its neighboring nodes, hence introducing a problem called idle-listening problem. This situation causes a node to expend a lot of wasteful energy causing the implementation of these protocols in WSN inefficient. Sensor-MAC (SMAC) protocol [8] attempted to solve the problems by introducing active-sleep cycles in the presence of random access channel. Node will execute a variant of MACA contention-based MAC protocol during active period to minimize the hidden terminal problem, while turning its radio off during sleep period to reduce idle listening problem. Furthermore, SMAC implements neighbors information variables called Network Allocation Vector (NAV) [9] for its collision avoidance technique. Node checks the NAV value before sending the RTS message. Nevertheless, implementing contention based mechanism is still vulnerable to collision due to random mechanism in its data packet transmission. Energy inefficiency caused by the idle-listening problem and high collision probability can be avoided in Time Division Multiple Access (TDMA) based protocols. In TDMA-based protocol such as HiperLan-II [10], time is divided into several frames, and a frame is divided into a number of time slots. Since all transmissions within the frame are pre-scheduled, it is possible for a node to sleep when it is not expected to transmit or receive any packets. Thus, the TDMA-based MAC protocol can clearly avoid the over-emitting problem. Since only the owner of the time slot is allowed to transmit a packet, collision problem can be avoided significantly. However, the dependency on a centralized base station implemented in HiperLan-II is not desirable for ad hoc deployed sensor networks. Therefore, the design goal should be to develop a TDMA-based protocol with distributed time slot scheduling algorithm. III. ENHANCED LIGHTWEIGHT MEDIUM ACCESS CONTROL Enhanced Lightweight Medium Access (el-mac) control inherits the good features of Lightweight Medium Access Control (LMAC) introduced in [11]. The enhancements include an active/sleep mechanism with predefined duty cycle for more efficient energy usage, reduced states for time slot assignment and an adaptive multi-timeslot allocation. LMAC can be categorized as a scheduled based or time division multiple access-based mechanism where time is structured into a number of fixed length time slot as shown in Fig 1. The time slot is further divided into two sessions; beacon session and data session. Each node in the network has its own time slot, allocated through its neighborhood information. Node transmits a beacon packet in the beacon session purposely to broadcast its own status to the neighboring nodes. This method is called distributed time slot assignment which differs from traditional time slot assignment, such as piconet in Bluetooth protocol and IEEE 802.15.4 protocol in beacon mode, where the latter requires a coordinator to assign a time slot to each node in the same cluster. 297

Time slot n-2 Time slot n-1 Time slot n Figure 1 Frame structure for el-mac protocol The structure of the beacon message is shown in Table 1 which has a length of 8 bytes. The time slot field will show the current time slot sequence followed by the level field which is used for time synchronization. The occupied time slot field is used to inform neighboring nodes which time slot to be used and which is free. The size of this field is 4 bytes or 32 bits length where the value for each bit indicates the availability of time slots. The last field in the beacon message is destination node address which shows the intended receiver in the next data session. Table 1 Beacon message structure Field t slot t bcn t listen t sleep Size (byte) Time slot 1 Level 1 Occupied time slot 4 Destination Node 2 Node keeps the information of its neighboring nodes by updating it in the neighborhood table and the format of the table is illustrated in Table 2. Table 2 Fields in neighborhood table Field Size (byte) Node address 2 Node Level 1 Node time slot 1 Data Sequence Number 1 Node Transmit Level 1 A. el-mac Distributed Time slot Allocation Unlike classic TDMA mechanism which uses a coordinator to assign child node s time slot, the el-mac protocol inherits the distributed time slot assignment mechanism from LMAC protocol with some simplification. In LMAC protocol, the process of obtaining a time slot by a sensor node is divided into four state; initial state, waiting state, discovery state and active state [11]. However, el-mac differs from LMAC t protocol whereby the waiting state and the discovery state have been combined as illustrated in Fig. 2. When a new node enters a network, it starts the algorithm in initial state where node listens to the channel in order to find a beacon signal from its neighborhood. Node turns into wait and discover state when it receives a beacon signal which enables it to start frame synchronization. Node shall wait for a random frame delay before selecting a free time slot and at the same time it shall also discover its neighboring nodes status by collecting the information from the received beacon signal. At the end of this state, node shall use Equation 1 to find any unoccupied time slot. z OR ( x1, x2,..., xn ) = x1ν x2ν... ν xn (1) Get a synchronization beacon message Wait and Discover Initial Lost synchronization beacon message Finish waiting and get a free Lost synchronization beacon and can t find other synchronization node Active Figure 2 State diagram for distributed time slot assignment in el-mac In equation 1, the value of x i is the i th occupied time slot collected from the beacon message while v is an OR operation. The resultant bitmap pattern from this operation will be in terms of 1 s and 0 s where 1 s will indicate occupied time slot, while 0 s indicate the vacant ones. Hence, a node randomly chooses its time slot identification from the list of vacant ones (indicated by 0 s). Bitmap vector 01100010 6 1 01010010 x 01101010 3 2 01011010 01100010 00101010 01011000 Figure 3 Distributed time slot assignment in el-mac Fig 3 shows an example of a network consisting of seven nodes operating in active mode and one new node (node x). The bitmap vector consists of 8 bits with the most significant bit corresponding to time slot 7, while time slot 0 is the least significant bit. Using equation 1, 4 6 1 298

the vacant time slots are 0, 5, and 7. Therefore, node x will randomly select its time slot from the list of vacant time slots. Node will enter active state when it has successfully selected a time slot. In this state, node will continuously transmit a beacon message at the beginning of its time slot to indicate its allocated slot. It will also listen to the channel at the beginning of other time slots in order to be able to listen for a beacon message from its neighboring nodes. Node enters sleep mode in two scenarios. First, after transmitting a beacon message and no more data packet scheduled to be transmitted. Second, if it receives a beacon message from it neighboring node indicating no incoming data packet. A TDMA-based MAC protocol requires nodes in the network to synchronize with each other. For el-mac protocol, a hierarchical time slot synchronization scheme were implemented where node will synchronize its time slot timer to another neighboring node s beacon signal which has the lowest level. This information is gathered from beacon messages transmitted by each neighboring nodes. The base station or sink node of the network will start with level 0. The time slot synchronization mechanism is also robust to time drifting problem caused by imprecision clock system in microcontroller and unpredicted operating system condition in sensor node [12], which would feasibly cause misalignment on beacon transmitting and listening period. Listening node needs to wake up from sleep mode earlier than the expected beacon transmission which is known as guard period. IV. DEVELOPMENT OF EL-MAC TEST BED The test bed consists of two platforms which are the hardware platform and the software platform. The wireless sensor module from Crossbow, namely the TelosB is used as off-the-shelf hardware platform which has a processor, a radio transceiver and a number of sensors embedded in it. On the other hand, the software platform is based on TinyOS which is the resident Operating System (OS) in TelosB and also a Java-based programming language. TinyOS is used to develop the new el-mac protocol that will control the sensors while Java programming is used to develop the interaction between sensor nodes and users. A graphical user interface (GUI) is also developed in the base station using Java programming language. A. Telos B TelosB was developed by University of California at Berkeley[13] which uses a low power Texas Instrument MSP430 microcontroller [14] as the main controller. The controller will read sensor data and then packetizes it. All protocols including network and MAC are implemented from here. To cater for wireless communication, the platform uses Chipcon CC2420 radio transceiver [15] which complies to IEEE 802.15.4 standard. The transceiver is controlled by the microcontroller via a serial peripheral interface (SPI). Additional modules inside the platform are data logger, platform serial identification, LEDs with different colors, and an external connector. The platform also comes with three internal sensors; humidity sensor, temperature sensor and light sensor. Fig. 4 shows the block diagram of TelosB. B. TinyOS Figure 4 Block Diagram of TelosB TinyOS is an open source OS designed for wireless embedded sensor networks [16]. It features a componentbased architecture which enables rapid innovation and implementation while minimizing code size as required by the severe memory constraints in sensor nodes. The OS is categorized as an event driven OS where code execution is run based on the occurred event and hence, is less responsive compared to other types of OS. TinyOS uses NesC as the programming language which is a programming language for networked embedded system that represents a new design aspect for application developers. It creates a C executable code that provides all the low level features for accessing hardware resources [17]. The programming model in NesC incorporates event driven execution, flexible concurrency model, and component-oriented application design. The NesC compiler can perform whole program analysis, including data race detection which will improve system reliability. 299

NesC has two types of components: configurations and modules. Configurations are used to wire modules together, while modules allocate state and implement executable logic. Modules also provide interfaces to and from other modules. Interfaces consist of commands and events function are used to specify the relationship between modules. Commands are implemented by modules that provide the interface while events are implemented by modules that use the interface. The development of TinyOS application is done using Cygwin under Windows XP OS. Cygwin is a Linux-like environment for Windows OS. It consists of two parts: a Dynamic Link Library (DLL) which acts as a Linux Application Programming Interface (API) emulation layer that provides substantial Linux API functionality and a collection of tools [18]. C. Implementation of el-mac To implement el-mac protocol, development work is needed for the sensor module in all nodes and also development of the GUI at the base station. The base station will be the sink node where packets from all nodes are sent to this node. The process begins with development of the TinyOS source and is followed by compilation work of the source code. The generated hexdecimal file from compilation process is programmed into the sensor module through the USB port. The execution of the TinyOS application is then validated and if an error occurred, debugging process will clear out the error. The second phase of the test bed development is to develop the GUI for the base station. V. FIELD TEST RESULTS All 9 nodes were set up and arranged in a 3 X 3 grid topology with 8 edge nodes periodically sending a packet with a payload size of 25 bytes to the center node, 0. Refer to Fig. 5. The sink node is connected to the base station for the purpose of data logging the experimental data. 1 2 3 8 0 4 (a) (b) Figure 6 (a) Sensor node mounting (b) Connection of sink node to base station Performance of simulation work as reported in [20] and field test results were compared. The throughputs were tabulated as shown in Fig. 7. It can be seen that the throughput of the experimental result for eight contending nodes is 10% less than the simulation result. The smaller throughput is caused by unpredictable test bed environment which leads to the higher number of packet loss. The propagation model used in the simulation may not be exactly similar to the real environment where transmitted signal is affected by fading, reflection, diffraction and interference problems. The difference is also due to the captured time of packet arrival caused by two types of delay: sensor node to PC delay and radio transceiver to microcontroller delay. Sink node uses serial communication to transfer the received packet to the PC. The data transfer rate of this communication is 115.2 kbps which is about half of the data transfer rate of the radio transceiver. Hence, packets that arrived at the sink node need to be queued before being transferred to the PC, producing unpredictable delay. The second factor is caused by the delay of data transfer from radio transceiver to the microcontroller using SPI communication. In the simulation, the time of packet arrival is captured instantly when the packet arrives at a node. Hence, the captured time is more precise. 7 6 5 Figure 5 Network topology in simulation and test bed environment Sensor nodes are placed on a 1 meter high stand while the sink node, 0 is connected to the base station via the USB port as shown in Figure 6. The distance between each adjacent node is 10 meter 300

VII. ACKNOWLEDGEMENT This project was funded by the Malaysia Ministry of Science, Technology and Innovation e-science grant 01-01-06-SF0306. Figure 7 Comparison between experiment and simulation results for throughput Fig. 8 shows the performance of the experiment as compared to the simulation in terms of packet received ratio (PRR). The experimental test bed and simulation results are comparable with a slight degradation in the experimental results. Here, the packet loss is also caused by unpredictable environment condition while running the test bed. However, the PRR is still above 0.9 which is considered as a good ratio [19]. This is due to the scheduling algorithm applied in the el-mac protocol which allows only one packet transmission in a slot, hence, preventing two nodes from transmitting a packet simultaneously which may cause packet collision. Furthermore, it can be seen that although the packet generation rate is increased, the packet received ratio is still high. Figure 8 Comparison between experiment and simulation results for packet received ratio VI. CONCLUSION This paper has presented the implementation of el- MAC protocol in a real test bed which consists of 9 TelosB sensor nodes programmed in the TinyOS of the node. NesC programming language was used to implement the protocol in the WSN module. Even though, there is a degradation of the results compared to results obtained from simulation, the difference is minimal but consistent and reflects the real implementation environment. REFERENCES [1] A. Mainwaring, J. Polastre, R. Szewczyk, D. Culler, and J.Anderson. Wireless Sensor Networks for Habitat Monitoring. International Workshop on Wireless Sensor Networks and Applications. 2002, pp. 88-97. [2] D. Estrin, R. Govindan, J. S. Heidemann, and S. Kumar. Next Century Challenges: Scalable Coordination in Sensor Networks. Mobile Computing and Networking, 1999, pp. 263-270. [3] D. Snoonian. Smart Building. IEEE Spectrum Magazine, 40(8), 2003, pp. 18-23. [4] Andrew S. Tanenbaum. Computer Networking. Prentice-Hall, 2002. [5] N. Abramsom. The Aloha System Another Alternative for Computer Communication. AFIPS, 1970, pp. 265-298. [6] Kleinrock, L. and F. Tobagi. Carrier Sense Multiple Access for Packet Switched Radio Channels. Int. Conf. on Communications, June 1974, pp. 21B-1 to 21B-7. [7] P. Karn. MACA: A New Channel Access Method for Packet Radio. 9 th ARRL Computing Networks Conference, Sept. 1990, pp. 134-140. [8] W. Ye, J. Heidemann, and D. Estrin. Medium Access Control with Coordinated Adaptive Sleeping for Wireless Sensor Networks. IEEE/ACM Trans. Net., vol. 12, no. 3, June 2004, pp. 493 506. [9] IEEE Computer Society. Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specificatio. IEEE Std. 802.11, 1997. [10] ETSI. HIPERLAN Type 2 Technical Specification. DTS/BRAN- 0023003, Aug. 1998. [11] L V. Hoesel, P. Havinga. Collision-free Time Slot Reuse in Multihop Wireless Sensor Networks. Proc. Int. Conf. on Intelligent Sensor, Sensor Networks and Information Processing, Dec. 2005, pp. 101 107. [12] L. F. W. van Hoesel and P. J. M. Havinga. Design Aspects of An Energy-Efficient Lightweight Medium Access Control Protocol for Wireless Sensor Networks. July 17, 2006. [13] Crossbow Technology. Telosb: An IEEE 802.15.4 Compliant wireless node.http://www.xbow.com/products/product_pdf_files/wireless_ pdf/telosb_datasheet.pdf. Accessed on Apr. 1, 2007. [14] Texas Instrument. MSP430: 16-bit Ultra Low-power MCUs. http://www.ti.com/lit/gpn/msp430f2410. Accessed on May 20, 2007. [15] Chipcon. CC2420: Single-chip 2.4 Ghz IEEE 802.15.4 Compliant Transceiver. http://www.ti.com/lit/gpn/cc2420. Accessed on Apr. 1, 2007. [16] J. Hill, R. Szewczyk, A. Woo, S. Hollar, D. E. Culler and K. Pister. System Architecture Directions for Networked Sensors. In Architectural Support for Programming Languages and Operating Systems. 2000. pp. 93-104. [17] D. Gay, P. Levis, R. von Behren, M. Welsh, E. Brewer and D. Culler. The NesC Language: A Holistic Approach to Networked Embedded System. In Proc. ACM SIGPUN 2003, Conf. on Programming Language Design and Implementation. June 2003. pp. 1-11. [18] Cygwin Information and Installation. http://www.cygwin.com. Accessed on Mar. 23, 2007. [19] S. Lin, J. Zhang, G. Zhou, L. Gu, T. He and J. A. Stankovic. ATPC: Adaptive Transmission Power Control for Wireless Sensor Networks. In Proceedings of the 4th International Conf. on Embedded Networked Sensor Systems. Colorado. 2006. pp. 223-236. [20] Rozeha Abd Rashid, Wan Mohd Ariff Ehsan W Embong, Norsheila Fisal. Enhanced Lightweight Medium Access Control Protocol. Radio Frequency & Microwave Int. Conf. (RFM). KL, Dec 2-4, 2009 301