FAMA/TDMA Hybrid MAC for Wireless Sensor Networks

Transcription

1 FAMA/TDMA Hybrid MAC for Wireless Sensor Networks Nuwan Gajaweera Dialog-University of Moratuwa Mobile Communication Research Lab, University of Moratuwa, Katubedda, Moratuwa, Sri Lanka Dileeka Dias Department of Electronic & Telecommunication, University of Moratuwa, Katubedda, Moratuwa, Sri Lanka Abstract Wireless sensor networks are an emerging area of interest among embedded and wireless research groups around the globe. It is envisioned that these low power computers will become the enabling technology in pervasive computing and bring about the next paradigm shift in computing. We present a new type of MAC protocol that is applicable to a special type of mobile wireless sensor network. In this network, mobile sensor nodes will sense the environment, and buffer the measured data for later retrieval. Once within range of the base station, the mobile mote will have a considerable amount of data to upload to the base station. Such a system could be used in different applications such as analyzing migration patterns of animals (ex. Elephants) or offline monitoring of vehicle fleets. We present a MAC protocol that will maximize the throughput of the said network, and thereby ensure that the maximum amount of data can be uploaded by the mobile mote to the base station in a given time. By using a combination of floor acquisition multiple access (FAMA) and TDMA an optimal MAC protocol for the above mentioned scenario is designed. To gain access to the base station, motes will first contend with each other by sending RTS messages to the base station using a CSMA scheme. The mobile node with the first successful RTS message will be registered with the base station. We achieve high channel utilization by providing a single node with unmitigated access to the medium for the duration of the upload i.e. the mote will acquire the floor. Uplink data transfer is carried out using a TDMA like scheme, where every time slot is allocated to the said mote. The proposed protocol is implemented as TinyOS components, targeted for the MICA2 sensor network platform. Simulation results are presented that benchmark the proposed FAMA/TDMA Hybrid MAC against a CSMA based MAC (B- MAC) protocol as well as a demand assigned TDMA protocol. I. INRTODUCTION Wireless sensor networks of tomorrow, as the enabling tool for pervasive ubiquitous computing will be called upon to operate in diverse scenarios. The classic example of wireless sensor network (WSN) deployment is the building monitoring application. Here the sensor nodes may be deployed in high densities to monitor and adjust temperature, humidity, light level, etc. But due to its inherently versatile nature, these devices may find their way into diverse application scenarios never before imagined. We present a new type of media access control (MAC) that will be the enabling technology in a special type of mobile sensor network. In this type of network the mobile motes will sense the environment, and buffer the measured data for later analysis, while moving in and out of radio range of a base station. Once within range the motes will upload the buffered data to the base station. Possible applications could range from analyzing migratory patterns of animals (ex. elephants) to offline monitoring of vehicle fleets. In either case the motes should be mounted on the target or object of interest (ex. vehicle or animal under study) II. BACKGROUND Much research has been carried out on network protocols for deeply embedded wireless sensor networks. These protocols can be thought of as lightweight versions of network protocols found on more established wireless networks such as cellular networks and wireless LANs. What distinguishes the network protocols found on WSN from other wireless networks protocols is the focus on conservation of energy. This is due to the fact that sensor nodes are expected to run off batteries, and are expected to be deployed in a manner that prohibits replacement or replenishment of these batteries. If we were to narrow our focus on MAC protocols for wireless sensor networks, one would observe that researches have focused mainly on three types of protocols [1]. These are random access schemes, slotted schemes and time division multiple access (TDMA) based schemes. Additionally hybrids of these schemes also exist. Examples of random access based MAC protocols for sensor networks range from Aloha variants [2] to carrier sense multiple access (CSMA) based schemes. A commonality between all these protocols is preamble sampling technique [2], [3]. Transmitting nodes will send out a long preamble before each data packet is sent out. Nodes will periodically wake up from sleep mode and will sample the medium for a preamble. If a preamble is detected the said node(s) will stay awake to receive the packet. Berkley-MAC (B-MAC) [3] is a popular CSMA based MAC protocol, which also happens to be the default MAC protocol in TinyOS [4]. Two attractive features of B-MAC (as in many other CSMA based schemes) are its flexibility (can be adapted to any network topology) and scalability /08/$ IEEE ICIAFS08

2 Sensor-MAC (S-MAC) [5] and Timeout-MAC (T-MAC) [6] and are classified as slotted schemes. Here the nodes employ periodic sleep/active cycles. The node will carry out radio communication only during the active period. Thus each node has to learn and maintain recodes of its neighbor s sleep/wakeup schedule. In effect these protocols trade in delay and throughput in order to save energy. Although TDMA MAC protocols offer much promise as WSN MAC protocol of the future, a flexible and scalable TDMA based MAC is yet to be developed. One reason may be the requirement for a base station/access point that restricts the formation of ad-hoc networks. PEDEMACS (Power Efficient and Delay Aware Medium Access Protocol for Sensor Networks) [7] is an early attempt at a TDMA based WSN MAC. Another popular protocol in this category is LEACH (Low Energy Adaptive Clustering Hierarchy) [8]. Unlike the earlier protocols LEACH is a routing protocol, which portions the network in to clusters. Nodes within the cluster communication in a TDMA fashion, while cluster heads form among themselves a tree network with the gateway/base station at the root. Although TDMA based schemes are inflexible by nature, it is clear that given a specific type of network, it is possible to design a TDMA based protocol that has been optimized for that type of network. As mentioned earlier, many of the WSN MAC protocols were inspired by wireless MAC protocols that were designed for other wireless networks. Similarly this work also draws it s inspiration from Floor Acquisition Multiple Access (FAMA) [9], a MAC protocol designed for wireless LANs FAMA is a multiple access scheme which has some resemblance to MACA (Medium Access with Collision Avoidance) [10] that is presently used in wireless LANs. MACA avoids the hidden node problem by the exchange of RTS-CTS packets. The RTS-CTS packet exchange reserves the channel for the single data packet that follows i.e. the channel (or floor) is acquired on a packet by packet basis. FAMA takes this a step further by acquiring the channel (floor) for a number of packets. The authors of FAMA argue that FAMA throughput is comparable to that of non-persistent CSMA but at the same time avoids the hidden terminal problem. III. THE APPLICATION SECNARIO As stated earlier, the FAMA/TDMA MAC was designed for a specific type of mobile sensor network. This network comprises of mobile sensor nodes (motes) and a stationary base station node. The mobile motes will sense the environment (e.g. GPS coordinates), and log this information to an onboard nonvolatile memory. When within radio range of the base station the mobile motes will upload the logged data to the base station. It therefore becomes apparent that the actual data communication is predominantly from the mote to the base station (uplink), while down link traffic may be limited to control traffic. Since the motion of the motes is random, it is quite possible that motes may move in and out of radio range of the base station. Therefore the MAC protocol should maximize the throughput in order to retrieve as much data as possible when the mote is within radio range. From the perspective of the mote, data transfer periods are limited to short bursts, and therefore conservation of energy is not a crucial requirement as in other wireless sensor networks. On the other hand energy conservation on the base station is not required, since the device will permanently be connected to a power supply. Additionally, multiple motes may be present at the base station and may wish to communication with the base station. Therefore the developed MAC protocol also has the job of allocating the bandwidth fairly among contending motes IV. DEMAND ASSIGNED TDMA MAC As precursor to the FAMA/TDMA MAC, a demand assigned TDMA MAC protocol was designed. The network is envisaged to be comprised of a single hop network containing a single base station and multiple mobile motes that wish upload data to the base station. Time is divided into contention periods where motes will contend to register with the base station, and transmission periods where the actual data transmission takes places in TDMA fashion. Transmission and contention period start and end messages are broadcast by the base station. Transmission and contention periods are time interleaved. Messages exchanged during the contention period are listed in Table I. TABLE I. Message Type Contention period start RTS Contention period end MESSAGES EXCHANGED DURING THE CONTENTION PERIOD Direction Downlink Uplink Downlink Description Signals the start of a new contention period Uplink packet sent by the mote to request entry into the tx group. Signals the end of the contention period. This packet also serves as an acknowledgement of an RTS packet. If RTS by a mote was accepted by the base station, the address of that particular mote will also be included in this packet. The contention period is advertised by the base station by sending the contention period start message, following which motes will send RTS messages to the base station. The first mote to send a successful RTS message is registered with the base station. This mote becomes a member of the transmission group i.e. the set of motes that are allowed to transmit during the transmission period.

3 Messages exchanged during the contention period are listed in Table II. TABLE II. MESSAGES EXCHNAGED DURING THE TRANSMISSION PERIOD Message Type Direction Description Transmission Signals the start of a new contention Downlink period start period Data Uplink Uplink packet sent by the mote to request entry into the tx group. Signals the end of the contention period. This packet also serves as an Upload End Downlink acknowledgement of an RTS packet. If RTS by a mote was accepted by the base station, the address of that particular mote will also be included in this packet. Signals the end on of a transmission period. In addition this packet also Contention Downlink contains acknowledgements for the data period end packets correctly received during the tx period The transmission period is advertised by the base station by sending the transmission period start message, following which members of the transmission group will send data packets in the allotted slots. It should be noted that the number of time slots in a transmission period is fixed. When there is a change in the transmission group membership i.e. a new mote has joined the group or an existing one has left the base station will piggy-back time slot allocation information with the transmission period start message. If the number of transmission group members is less than the number of time slots, one mote will be allowed to send multiple packets during a single transmission period. The base station will piggy-back acknowledgement information with the transmission period end message. A packet error on the down link will be detrimental to communication. Therefore the base station will transmit at a high power level to reduce the chance of packet errors. Motes will transmit at a lower power level. The base station will monitor the received power level of RTS messages. If the received power level of an RTS packet is below a certain threshold, the requesting mote will not be admitted into the transmission group. This selection procedure bars entry to motes that are further away, and will thus reduce packet errors on the down link. In the above mentioned protocol, overhead is mainly incurred due to the following two reasons; Frequent contention periods if the number of motes in the transmission group is less than the maximum number of time slots, the base station will advertise contention periods. Each time the membership of the transmission group changes an uplink map is sent out. In simulations carried out, it was observed that the protocol is efficient only when the number of nodes were high, whereas poor results were observed for lower node counts. V. FAMA/TDMA HYBRID MAC With the above realization a new approach was adopted that would eliminate the above mentioned drawback, but at the same time maintain the benefits of a TDMA system. Simply put, it was decided to limit the transmission group membership to one. This effectively means that the medium is acquired by a single node for the transmission of multiple packets i.e. the floor is acquired by the registered mote. Hence the resulting protocol may be viewed as a hybrid between FAMA and TDMA. This change has the following consequences; Infrequent contention period advisements a contention period would only be advertised when the registered mote has uploaded all its data Elimination of the uplink map since there is only one mote, an uplink map is not required. The MAC protocol has maximum channel utilization, regardless of the number of nodes in the networks To ensure that the same mote doesn t acquire the medium in two consecutive sessions, once a mote deregisters with the base station, it will stop all radio related activity for a fixed duration of time. VI. PROTOCOL DELAY For the MAC protocols presented here, protocol delay can be defined as the amount of time that a mote has to wait before it is granted access to a base station i.e. mean delay to register with the base station. The factor that determines the protocol delay is the frequency of contention period advertisements. As such the delay of the FAMA/TDMA Hybrid MAC would be grater than the delay of the demand assigned TDMA MAC, since the frequency of contention period advertisements is grater in the second protocol. Although the delay is grater in the FAMA/TDMA protocol, since only one mote is transmitting, it will take a lesser amount of time to upload all its data as compared with the demand assigned TDMA MAC. By adjusting the maximum amount of data that a mote is allowed to upload during a session, the delay can be brought down to a tolerable level. VII. DEVELOPMENT ENVIRONMENT TinyOS [4] is a component based event driven operating system that has been designed for wireless sensor networks. It has been widely adopted as a research platform by many wireless sensor network research groups. TinyOS applications are actually an interconnected set of components that communicate with each other using clearly defined interfaces. Each TinyOS component is in itself a state machine. TinyOS has been ported to a number of processors and platforms. Although most of the TinyOS components are reusable, some components such as the MAC protocol that directly

4 communicate with the hardware are platform specific. The MICA2 [11] is a wireless sensor node hardware platform this is supported by TinyOS. Due to the above mentioned reason the MICA2 was selected as the platform on which to develop and evaluate the above mentioned MAC protocols. The device comprises of an ATMega128 [12] microcontroller and a Chipcon CC1000 [13] radio transceiver. Due to limited availability of MICA2 hardware, and ease of development, a simulation environment was selected on which easy evaluation of the developed network protocols could be carried out. The Avrora simulator was selected for this purpose. Avrora [14] is cycle-accurate simulator of the AVR microcontroller, allowing real programs to be run with precise timing. In addition to this Avrora is able to accurately simulate a network of MICA2 motes. Thus any code written for and tested on Avrora can be directly run on an actual MICA2 network. VIII. SIMULATION Simulations were first carried out using Avrora, and the obtained results were then compared against results from a deployment of the protocol on real hardware. For the purpose of the simulation, a single hop network where all the nodes were within radio range of one another was considered. In line with the requirement of the actual network, during the simulation motes would attempt to upload the maximum amount of data to the base station in a given time. The following network protocols were evaluated; B-MAC a CSMA MAC protocol for TinyOS FAMA/B-MAC Hybrid Demand Assigned TDMA MAC FAMA/TDMA Hybrid MAC The requirement was to assess which protocol would maximize the radio resource usage. As such the overall throughput i.e. the rate at which data arrives at the base station was measured. A. Pure B-MAC Simulation B-AMC is the default CSMA based MAC that ships with TinyOS. The Simulations were carried out using Avrora for different number of motes as well as mote ID combinations. In the simulated mote network, one mote was designated as the data sink (base station), while the other motes (data sources) would contend with each other to send data to the base station. To introduce randomness to the simulator, motes were made to wake up after a random delay. Since the random number generator in TinyOS (which is also used by B-MAC), is initialized using the node ID, several simulation runs were carried out using different (randomly selected) node id combinations. B. FAMA/B-MAC Simulation Here a floor acquisition multiple access scheme (FAMA) was built on top of B-MAC as an application level component. In this scenario, the base station would send an invitation to motes to register with the base station, following which motes would content with each other to register. The successfully registered mote would then be allowed to transfer a fixed number of packets termed packet per cycle after which the mote would deregister with the base. Simulations were carried out for different node ID combinations for packets-per-cycle (PPC) values of 32, 64, 128 and infinity. C. Demand Assigned TDMA MAC The demand assigned MAC protocol described in Section IV was implanted in TinyOS. Simulations were carried out using a 12 slot TDMA frame. Similar to the above scenario members of the transmission group were allowed to transfer a fixed number of packets before deregistering with the base station. Simulations were carried out for different node ID combinations for packets-percycle (PPC) values of 32, 64, 128 and infinity. D. FAMA/TDMA Hybrid MAC The FAMA/TDMA MAC protocol described in Section V was implanted in TinyOS. Simulations were carried out using a 12 slot TDMA frame. The registered mote was allowed to transfer a fixed number of packets before deregistering with the base station. Simulations were carried out for different node ID combinations for packets-per-cycle (PPC) values of 32, 64, 128 and infinity. Throughput (η) inf PPC 128 PPC 64 PPC 32 PPC BMAC IX. SIMULATION RESULTS A. B-MAC & FAMA/B-MAC Simulation The simulation results for the Pure B-MAC and FAMA/B- MAC Hybrid protocols are given in Fig Node Count (N) Figure 1. Throughput vs. Node Count for the Pure B-MAC and FAMA/B- MAC Hybrid protocols As can be seen, although the performance of B-MAC degrades as the number of nodes increase, the performance of the hybrid protocol is fairly consistent.

5 It should be noted that the throughput at infinite packets per cycle was measured to get an idea on the theoretical best performance of the protocol, and has no practical use. B. Demand Assigned TDMA MAC As depicted in Fig. 2 the TDMA MAC shows better performance than B-MAC as the number of nodes increase. D. Simulation verification In order to verify the simulation environment, a similar scenario was set up in the real world using actual MICA2 motes. Due to the limited availability of hardware motes, the setup included one mote as the base station and two motes as mobile nodes. The obtained results showed good agreement with the simulation output. Throughput (η) Throughput (η) inf PPC 128 PPC 64 PPC 32 PPC Node Count (N) Figure 2. Throughput vs. Node Count for the demand assigned TDMA MAC In contrast to the FAMA/B-MAC hybrid, where the throughput decreased with the number of nodes, here the throughput increases with the number of nodes. At lower node numbers, the base station frequently advertises contention periods. Therefore the overhead is greater at lower node counts, hence low throughput. C. FAMA/TDMA Hybrid MAC The simulation results for the Pure B-MAC and FAMA/B- MAC Hybrid protocols are given in Fig. 3. inf PPC 128 PPC 64 PPC 32 PPC Node Count (N) Figure 3. Throughput vs. Node Count for the FAMA/TDMA hybrid MAC It is clear from the above that the performance of the FAMA/TDMA MAC is far superior to all other MAC protocols for the given scenario. In addition to this the throughput is almost constant at any node count. This indicates that the available radio resources are utilization to the maximum at all times. X. CONCLUSION In this paper we have presented a MAC protocol for a special type of wireless sensor network i.e. a single hope network where motes upload a significant amount of data to a base station. By combining two very different multiple access schemes a novel type of MAC protocol has been designed that would maximize the available radio resource usage. By combining FAMA and TDMA the designed protocol provides contention free access to a single mote. Although in the short term, this may result in unfair allocation of radio resources, it becomes apparent that in the long run each mote in the network is allowed fair access to the medium. Additionally by providing contention free access to the medium for a single mote, the FAMA/TDMA Hybrid MAC is able to achieve a high channel utilization figure compared to other MAC protocols. Although the FAMA/TDMA Hybrid MAC may not be suitable for any type of wireless sensor network, for the scenario under question, it is by far the optimal MAC protocol. ACKNOWLEDGMENT The authors are grateful to Dialog Telekom PLC., Sri Lanka, for their support of this work. REFERENCES [1] K.G. Langendoen, 2006, June, The MAC Alphabet soup [online]. Available: [Accessed: Sep. 12, 2008]. [2] A. El-Hoiydi, Aloha with preamble sampling for sporadic traffic in ad hoc wireless sensor networks, in Proceedings of IEEE International Conference on Communications, vol. 5, pp , Apr [3] J. Polastre, J. Hill and D. Culler, Versatile low power media access for wireless sensor networks, in Proceedings of the 2nd international conference on Embedded networked sensor systems, 2004, pp [4] The TinyOS Alliance, TinyOS, an open-source OS for the networked sensor regime, 2007 [online], Available: [Accessed: Sep. 12, 2008]. [5] W. Ye, J. Heidemann and D. Estrin, An energy-efficient MAC protocol for wireless sensor networks, in Twenty-First Annual Joint Conference of the IEEE Computer and Communications Societies, 2002, pp [6] T. van Dam and K. Langendoen, An adaptive energy-efficient MAC protocol for wireless sensor networks, in Proceedings of the 1st international conference on Embedded networked sensor systems, 2003, pp [7] S. Coleri-Ergen and P. Varaiya, PEDAMACS: power efficient and delay aware medium access protocol for sensor networks, IEEE Transactions on Mobile Computing, vol. 5, pp , July [8] W. Heinzelman, A. Chandrakasan and H. Balakrishnan, Energy- Efficient Communication Protocol for Wireless Microsensor Networks, in Proceedings of the 33rd Hawaii International Conference on System Sciences, vol. 8, 2000, p

6 [9] C. L. Fullmer and J.J. Garcia-Luna-Aceves, Floor Acquisition Multiple Access (FAMA) for Packet-Radio Networks, in Proceedings of ACM SIGCOMM 95, 1995, pp [10] P.Karn, MACA a new channel access method for packet radio, in ARRL/CRRL Armature Radio 9th Computer Networking Conference, pp , Sep [11] Crossbow Technology, Inc., MICA2 Wireless Measurement System [online], Available: Wireless_pdf/MICA2_Datasheet.pdf [Accessed: Sep. 12, 2008]. [12] Atmel Corporation, ATmega128 8-bit Microcontroller with 128K Bytes In-System Programmable Flash [online], Available: [Accessed: Sep. 12, 2008]. [13] Texas Instruments, CC1000 Single Chip Ultra Low Power RF Transceiver for 315/433/868/915 MHz SRD Band [online], Available: [Accessed: Sep. 12, 2008]. [14] UCLA Compilers Group, Avrora The AVR Simulation and Analysis Framework [online], Available: [Accessed: Sep. 12, 2008].