AMAC: Traffic-Adaptive Sensor Network MAC Protocol through Variable Duty-Cycle Operations

Similar documents
SENSOR-MAC CASE STUDY

Implementation of an Adaptive MAC Protocol in WSN using Network Simulator-2

Presented by: Murad Kaplan

Research Article MFT-MAC: A Duty-Cycle MAC Protocol Using Multiframe Transmission for Wireless Sensor Networks

Performance and Comparison of Energy Efficient MAC Protocol in Wireless Sensor Network

Analysis of S-MAC/T-MAC Protocols for Wireless Sensor Networks

COMPARISON OF TIME-BASED AND SMAC PROTOCOLS IN FLAT GRID WIRELESS SENSOR NETWORKS VER VARYING TRAFFIC DENSITY Jobin Varghese 1 and K.

MAC LAYER. Murat Demirbas SUNY Buffalo

Reducing Inter-cluster TDMA Interference by Adaptive MAC Allocation in Sensor Networks

Reservation Packet Medium Access Control for Wireless Sensor Networks

An Energy-Efficient MAC Protocol for Delay-Sensitive Wireless Sensor Networks

CSMA based Medium Access Control for Wireless Sensor Network

CSC8223 Wireless Sensor Networks. Chapter 5 Medium Access Control Protocols

CS 410/510 Sensor Networks Portland State University

Wireless Sensor Networks 8th Lecture

Enhanced Power Saving Scheme for IEEE DCF Based Wireless Networks

WIRELESS sensor networking is an emerging technology

An Energy-Efficient MAC using Dynamic Phase Shift for Wireless Sensor Networks

Delay Analysis of ML-MAC Algorithm For Wireless Sensor Networks

Embedded Internet and the Internet of Things WS 12/13

AN EFFICIENT MAC PROTOCOL BASED ON HYBRID SUPERFRAME FOR WIRELESS SENSOR NETWORKS

Impact of IEEE MAC Packet Size on Performance of Wireless Sensor Networks

MAC in /20/06

AN ADAPTIVE ENERGY EFFICIENT MAC PROTOCOL FOR WIRELESS SENSOR NETWORKS

Medium Access Control (MAC) Protocols for Ad hoc Wireless Networks -IV

A Survey on Medium Access Control Protocols based on Synchronous Duty Cycle Approach in Wireless Sensor Networks

A MAC Protocol with Little Idle Listening for Wireless Sensor Networks

An Energy Consumption Analytic Model for A Wireless Sensor MAC Protocol

MAC Essentials for Wireless Sensor Networks

Power Saving MAC Protocols for WSNs and Optimization of S-MAC Protocol

QoS Challenges and QoS-Aware MAC Protocols in Wireless Sensor Networks

AN EFFICIENT MAC PROTOCOL FOR SUPPORTING QOS IN WIRELESS SENSOR NETWORKS

An Adaptive MAC Protocol for Efficient Group Communications in Sensor Networks

Advanced Networking Technologies

ABSTRACT. Physical Implementation of Synchronous Duty-Cycling MAC Protocols: Experiences and Evaluation. Wei-Cheng Xiao

UNIT IV. Data link layer protocols. Prof.Prasad S.Halgaonkar

Keywords T MAC protocol, reduction function, wsn, contention based mac protocols, energy efficiency; Fig 1. Listen and sleep cycle in S MAC protocol

R-MAC: An Energy-Efficient MAC Protocol for Underwater Sensor Networks

Outline. MAC (Medium Access Control) General MAC Requirements. Typical MAC protocols. Typical MAC protocols

Medium Access Control in Wireless Sensor Networks

End-To-End Delay Optimization in Wireless Sensor Network (WSN)

FTA-MAC: Fast Traffic Adaptive energy efficient MAC protocol for Wireless Sensor Networks

Improving IEEE Power Saving Mechanism

Low Power and Low Latency MAC Protocol: Dynamic Control of Radio Duty Cycle

Interference avoidance in wireless multi-hop networks 1

Etiquette protocol for Ultra Low Power Operation in Sensor Networks

EX-SMAC: An Adaptive Low Latency Energy Efficient MAC Protocol

An Enhanced Cross-Layer Protocol for Energy Efficiency in Wireless Sensor Networks

sensors ISSN

Chapter 3: Medium Access Control in Wireless Sensor Networks

Smart Hybrid Frame Scheduling to Improve Energy Efficiency in Wireless Sensor Network

Energy Efficient MAC Protocols Design for Wireless Sensor Networks

Collision Free and Energy Efficient MAC protocol for Wireless Networks

MAC Protocols 10/6/2008. References. Medium Access Control (MAC)

COMPARISON OF CSMA BASED MAC PROTOCOLS OF WIRELESS SENSOR NETWORKS

Ad hoc and Sensor Networks Chapter 5: Medium access control protocols

[Kaplieswar*, 5(4): April, 2016] ISSN: (I2OR), Publication Impact Factor: 3.785

MAC protocols. Lecturer: Dmitri A. Moltchanov

Power-efficient Communication Protocol for Social Networking Tags for Visually Impaired

Computer Communication III

An Efficient Medium Access Control Protocol with Parallel Transmission for Wireless Sensor Networks

Networking Sensors, I

Medium Access Control in Wireless Sensor Networks

Lecture 9: Bridging. CSE 123: Computer Networks Alex C. Snoeren

Ferry Route Design with MAC Protocol in Delay Tolerant Networks

An adaptive energy-efficient and low-latency MAC for tree-based data gathering in sensor networks

Micro-Frame Preamble MAC for Multihop Wireless Sensor Networks

CSE 461: Wireless Networks

Computer Networks 53 (2009) Contents lists available at ScienceDirect. Computer Networks. journal homepage:

RMAC: A Routing-Enhanced Duty-Cycle MAC Protocol for Wireless Sensor Networks

Medium Access Control in Wireless IoT. Davide Quaglia, Damiano Carra

Intra and Inter Cluster Synchronization Scheme for Cluster Based Sensor Network

TMMAC: A TDMA Based Multi-Channel MAC Protocol using a Single. Radio Transceiver for Mobile Ad Hoc Networks

An Energy-Efficient MAC Design for IEEE Based Wireless Sensor Networks

Maximizing the Lifetime of Clustered Wireless Sensor Network VIA Cooperative Communication

Principles of Wireless Sensor Networks. Medium Access Control and IEEE

RT-Link: A global time-synchronized link protocol for sensor networks Anthony Rowe, Rahul Mangharam, Raj Rajkumar

Medium Access Control in Wireless Networks

Strengthening Unlicensed Band Wireless Backhaul

Priority-MAC: A Priority based Medium Access Control solution with QoS for WSN

Rahman 1. Application

ROUTING ALGORITHMS Part 2: Data centric and hierarchical protocols

A More Realistic Energy Dissipation Model for Sensor Nodes

IRI-MAC: An Improved Receiver Initiated MAC Protocol for Wireless Sensor Network

Latency and Energy Efficient MAC (LEEMAC) Protocol for Event Critical Applications in WSNs

Research Article An Energy and Latency Aware WSN MAC Protocol for Bidirectional Traffic in Data Collection

We are IntechOpen, the world s leading publisher of Open Access books Built by scientists, for scientists. International authors and editors

Survey of Asynchronous Medium Access Protocols for Wireless Sensor Networks

Energy and Rate based MAC Protocol for Wireless Sensor Networks

Energy Management Issue in Ad Hoc Networks

A Directional MAC Protocol with the DATA-frame Fragmentation and Short Busy Advertisement Signal for Mitigating the Directional Hidden Node Problem

Delay Sensitive and Longevity of Wireless Sensor Network Using S-MAC Protocol

Energy Consumption and Fault Tolerance in the MAC Protocols for WSN

Exploiting Routing Redundancy using MAC layer Anycast to Improve Delay in WSN

Energy Management Issue in Ad Hoc Networks

CLUSTER-BASED ENERGY-EFFICIENT MAC PROTOCOL FOR WIRELESS SENSOR NETWORKS. A Thesis by. Nikhil Marrapu

Improving the Data Scheduling Efficiency of the IEEE (d) Mesh Network

MAC Protocols for Energy Conservation in Wireless Sensor Network

ViTAMin: A Virtual Backbone Tree Algorithm for Minimal Energy Consumption in Wireless Sensor Network Routing

S-MAC Protocol for Wireless Sensor Network and Study of related work

Transcription:

AMAC: Traffic-Adaptive Sensor Network MAC Protocol through Variable Duty-Cycle Operations Sang Hoon Lee, Joon Ho Park, and Lynn Choi Department of Electronics and Computer Engineering Korea University Seoul, Korea {smile97, cronide, lchoi}@korea.ac.kr Abstract Sensor network MAC protocols usually employ periodic sleep and wakeup, achieving low duty-cycle to save energy and to increase the lifetime of battery-powered sensor devices. However, existing protocols require all the sensor nodes to operate on the same static schedule, waking up all the nodes at the same fixed interval periodically. This paper proposes a new media access control protocol called AMAC that can achieve significant energy savings by dynamically changing the schedule of each node depending on the traffic. In AMAC, each node can adjust the duration of the periodic interval as well as the duration of the active period depending on the traffic. Thus, busy nodes can operate with a high duty-cycle while idle nodes can operate with a low duty-cycle at the same time, achieving both low-energy and high-performance. The results of our detailed simulations confirm that AMAC can reduce the average energy consumption by a factor of up to 6.8 compared to an existing fixed duty-cycle MAC protocol while it can also improve the network performance for burst traffic patterns. Keywords - sensor network, media access control, low-energy network protocol, duty-cycle, network performance I. INTRODUCTION Massive proliferation of various types of affordable mobile computing devices and wide adoption of wireless networking technologies has brought modern society the beginning of ubiquitous computing era. A vast number of small and low-cost sensors in capable of computing and wireless communication will eventually extend today s network of mobile devices into the true ubiquitous computing era in the foreseeable future. Designing communication protocols for wireless sensor networks face different challenges and constraints than existing wireless networks. While traditional MAC protocols assume nodes to wakeup all the time to maximize throughput, minimize latency, and provide fairness, waking up the nodes all the time is impossible in sensor networks since merely turning on the radio will soon deplete the node energy. Therefore, all the existing sensor network MAC protocols employ a low duty-cycle operation with a periodic sleep and wakeup either through a predetermined static scheduling of the channels [1, 2, 3], or through a coordinated synchronization of wakeup and sleep schedules [4, 5, 6]. This low duty-cycle operation can significantly save energy and extend the network lifetime at the expense of increased communication latency and synchronization overhead. However, existing sensor network MAC protocols suffer from either unnecessary energy wastes or performance penalty due to their homogeneous wakeup and sleep schedules. Most of them assume that all the nodes should wakeup at the same time synchronously to allow the sender to synchronize with any receiver [5, 7, 6, 8]. However, note that only a few nodes usually participate during communication. All the other nonparticipating nodes wake up unnecessarily, leading to idle listening. On one hand, increasing this periodic interval may further save energy at the cost of the increased communication delay and lower bandwidth. On the other hand, decreasing the periodic interval will improve the communication performance at the expense of increased energy consumption. In this paper we propose a new media access control protocol called AMAC. AMAC is fundamentally different from previously proposed protocols in that each node can dynamically adjust its own sleep and wakeup schedule. The idea is that nodes on the communication path should be on a high duty-cycle to maximize throughput and minimize latency while all the other nonparticipating nodes must operate on a lower duty-cycle to minimize energy consumption. The main ideas underlying AMAC are two-folds. First, each node can adjust the duration of the periodic interval dynamically depending on the network traffic. When a node is participating in the communication, the node can operate with a higher dutycycle by reducing this interval. When a node is idle for a long time, the node lowers its duty-cycle by increasing the interval. Thus, not all the nodes in the field wake up or sleep at the same time. Second, in addition to the variable periodic interval, a node can also adjust the duration of its active period dynamically depending on the traffic. A sender always broadcasts the communication token at the very beginning of an active period to inform the existence of communication. Thus, all listeners can adjust its active period by checking the communication token. Thus, when there is no traffic, a node can minimize its active period by removing unnecessary listen period for control packets. This dynamic adjustment of the active period as well as the dynamic adjustment of the periodic interval enables the duty cycle of each sensor node to adapt to the network traffic, resulting in significant energy savings for idle nodes and improved communication performance for busy nodes at the same time. We have implemented AMAC on the ns-2 simulation platform. The detailed simulation results confirm that AMAC can reduce the average energy consumption by a factor of up to 6.8 compared to an existing fixed duty-cycle MAC protocol while it can also improve the This research was supported by the MIC (Ministry of Information and Communication), Korea, under the ITRC (Information Technology Research Center) support program supervised by the IITA (Institute of Information Technology Assessment).

network performance for burst traffic patterns or for stream data types. The rest of this paper is organized as follows. Section II surveys the existing MAC protocols proposed for sensor networks, primarily focusing on contention-based schemes. In Section III we present the two main concepts of AMAC, namely the adjustment of periodic interval and the adjustment of active period, and describe its clock synchronization and schedule synchronization. In Section IV we present our simulation methodology and the detailed evaluation results of AMAC. Section V concludes the paper. II. RELATED WORKS MAC protocols can be classified into two broad categories: static allocation and dynamic allocation protocols. Both approaches have been studied in designing sensor network MAC protocols. However, the energy consumption issue, specifically minimizing the idle listening, has been the primary focus of the MAC protocol design rather than traditional design goals such as performance or fairness. Static channel allocation protocols allocate channels through a predetermined assignment such as TDMA, FDMA, CDMA, or combination of the three. Examples of sensor network MAC protocols in this category are SMACS [3], UNPF [1], and LEACH [2]. The main advantage of this approach is that it can provide a bounded delay. Among these static allocation protocols, TDMA protocols have a natural advantage of energy conservation, because the duty cycle of the radio is reduced and there is neither contention-induced overhead nor collisions. However, TDMA protocols are complex and difficult to implement in sensor networks because of lack of precise global clock synchronization. When the number of nodes is large and continuously varying, it is not easy for a TDMA protocol to dynamically change its frame length and time slot assignment. So its scalability is normally not as good as that of contention-based dynamic protocols. In addition, none of traditional static channel allocation methods such as TDMA or FDMA perform well with burst or variable traffic since most of channels will be idle most of time, bandwidth is lost and no one else is allowed to use it either. Dynamic allocation protocols allocate channels on demand. The advantage of this approach is its effectiveness in dealing with variable rate traffic, which is believed to be frequent in real-life scenarios of sensor network. This type of protocols can be further classified into reservation-based and contentionbased protocols. Reservation-based protocols reserve channels on demand. However, their large messaging overhead and long link setup delay make them not suitable for sensor networks. Contention-based protocols avoid the problems of reservationbased protocols by allocating channels based on random accesses rather than reservation. Most of recent protocols [5, 6, 8] can be classified into this category. Woo and Culler [9] examined different configurations of carrier sense multiple access protocols and proposed an adaptive rate control mechanism, which uses loss as a collision signal to adjust transmission rates in a manner similar to the congestion control in TCP. However, their main goal was to achieve fair bandwidth allocation to all nodes in a multi-hop network. In comparison, later schemes do not promote pernode fairness, and even trade it off for further energy savings. SMAC is one of the pioneering works in contention-based MAC protocols, specifically targeting sensor networks. SMAC proposes a periodic sleep and wakeup, providing a low duty cycle operation. To address the bandwidth and latency issues raised by the periodic wakeup, they also propose a message passing to send a long message of multiple packets in a single transaction by trading off the fragment-level fairness. TMAC [5] improves upon SMAC. In TMAC, the length of the active period is dynamically adapted to the traffic, using a timeout. The active period is ended whenever both physical and virtual carrier sensing find the channel idle for the given duration of the timeout. Both SMAC and TMAC assume all the nodes to wakeup simultaneously at the fixed interval. Recently a dynamic cycle time reduction technique called DSMAC [7] was proposed to address the performance issue related with periodic sleeping. However, DSMAC is different from AMAC in that the cycle time reduction is controlled by a sink node after the full communication path is established whereas in AMAC both cycle time reduction and extension are independently carried out by each node for both performance and energy optimizations. In addition, all the existing protocols derived from SMAC including TMAC and DSMAC, the duration of the active period includes the periods for RTS and CTS. However, when there is no traffic, such fixed RTS/CTS periods cause idle listening for all the nodes, which can be very substantial as demonstrated in Section IV. A rather different approach called preamble sampling was used in TinyOS project [10]. In the protocol, a sender is required to send a preamble long enough to cover the wakeup period of the receivers. This can not only minimize idle listening but it also allows each node to wake up asynchronously independent of other nodes. Thus, no networkwide clock synchronization is necessary. This idea is extended by WiseMAC [4]. In WiseMAC a sender can minimize the length of the preamble by exploiting the knowledge of the sampling schedules of its neighbors during communication, thus reducing the preamble transmission overhead. However, all the existing schemes based on the preamble sampling including WiseMAC suffer from the same fixed duty-cycle operation in that there is a dilemma between performance and energy depending on the length of its periodic interval. III. AMAC DESIGN We first define the terminology used in this paper. A cycle is defined as the periodic interval, which consists of an active period and a sleep period. An active period is further divided into SYNC and RTS parts as shown in Fig. 1(a). Each part has a contention window to reduce the chance of collisions. A cycle time is expressed as a multiple of the minimum cycle time T. The minimum cycle time T should be long enough to accommodate a single data transaction, i.e. the sequence of SYNC, RTS, CTS, DATA and ACK packets. T can be calculated from the size of each packet, the contention window size for each packet, transmission delay, and the RF transmission parameters of a given system.

A duty cycle in general is the proportion of time during which a component, device, or system is operated. In the context of this paper the duty cycle is defined as the ratio of the active period to the entire cycle time. The duty cycle can be expressed as a ratio or as a percentage. For example, a sensor node with a 1 second cycle time, which consists of 200 ms active period and 800 ms sleep period, is said to have a duty cycle of 20% or 0.2. after synchronization and a node can change its cycle time every T. A node A may change its cycle time from T to 2T at odd time units while another node B may change its cycle time from T to 2T at even time units. However, they can never synchronize each other. To avoid such a cycle skew, each node must change its cycle time on the natural cycle alignment boundary. For example, if a node with a cycle time of 8T wants to double its cycle time, it must change its cycle time at a multiple of 16T, i.e. 0, 16, 32 time units. This way, a faster node can always synchronize with slower nodes. For instance, a node with a cycle time of 2T is able to synchronize with the nodes with cycle times of 2T, 4T, and 8T, and so on. This is shown in Fig. 2(a). Figure 1. AMAC with adaptive active period. CS stands for carrier sense. (a) a busy receiver, (b) an idle receiver, and (c) a busy sender. A. Adaptive Active Period In AMAC, each node can dynamically adjust the duration of its active period depending on the traffic. The idea is that if there is no traffic in a cycle, we can remove the unnecessary RTS/CTS periods from the active period. To accomplish this, SYNC packet contains an extra bit field, called preamble bit, at the beginning of the packet. If the bit is set, then the SYNC packet tells all the listeners to extend their wakeup period to include RTS period since RTS and CTS will follow in this cycle. Thus, if a node has a data packet to transmit, it first constructs a SYNC packet by setting this preamble bit and broadcasts the SYNC packet to all of its listeners during SYNC period. This special SYNC packet is called communication SYNC while the regular SYNC packet is called clock SYNC. Thus, a receiver node first checks SYNC packet and if this preamble bit is set, the node extends its active period to listen for RTS packet. Otherwise, the node goes back to sleep immediately, reducing the duration of the active period and thus achieving lower duty cycle without any performance loss. Since there are two types of SYNC packets, they may compete for the channel during a SYNC period. Since clock synchronization is performed rarely, the chance of collision would be very small. However, during heavy traffic conditions, this might lead to a synchronization failure in the worst case. Thus, in our proposal, we gave the priority to the clock SYNC packet by padding a small carrier sense slots to the communication SYNC packet. B. Adaptive Cycle Time In AMAC each node can also adjust its cycle time depending on the traffic. This requires a careful coordination of the cycle time so that all nodes can easily synchronize with each other. For such coordination, in AMAC the cycle time of a node must be defined as a 2 n multiple of the minimum cycle time T. Thus, a node may have a cycle time of T, 2T, 4T, 8T, 16T, and so on. The maximum cycle time is determined by the latency requirement of the application running on the system. However, a cycle time may not be changed at any time to avoid schedule discrepancies. Assume that all nodes start at time 0 Figure 2. (a) Synchronization among the nodes with different cycle times and (b) dynamic cycle time adjustment. In AMAC, the decision of the cycle time adjustment is controlled by the traffic. If there is no traffic in the current cycle, all the listeners implicitly double their cycle time at their natural cycle alignment boundary to decrease its duty cycle during this idle period. This strategy is called automatic doubling, which is illustrated in Fig. 2(b). Thus a long idleness in the field would cause most of nodes to eventually operate with the maximum cycle time, leading to the maximum power saving mode. When an event occurs, the cycle time reduction is performed. A receiver needs to reduce its current cycle time to the minimum T to synchronize its active period with that of the sender if the node is specified as the receiver in the RTS packet. This strategy is called minimum reduction, which would lead to the highest performance. Reducing the cycle time gradually would increase the message latency without saving any more energy. C. Synchronization 1) Clock Synchronization Like SMAC, AMAC requires each node to synchronize its clock with those of its neighbors. This clock synchronization can be performed in two different approaches. As in SMAC we can employ a fully-distributed approach where all the nodes compete to become a SYNC packet originator. This is useful for a complete ad-hoc network scenario when no sink is available at the initial deployment stage. However, this is more time-consuming and can sometimes lead to multiple schedules with multiple virtual clusters [8]. Thus, nodes in the cluster boundary must wake up at the listen intervals of the two schedules, leading to more idle listening and overhearing.

Instead, we can take a centralized approach, where a sink initiates the synchronization by broadcasting the SYNC packet to its neighbors. Initially every node fully awakes until a SYNC packet is received from one of its neighbors before it starts periodic listen and sleep. As soon as it receives the SYNC packet, it synchronizes its timer and follows the same schedule as specified in the SYNC packet and announces its schedule and timing information at its next wakeup period. Each node hearing the SYNC packet constructs its own SYNC packet and rebroadcast to its immediate neighbors. Broadcasting a SYNC packet follows the normal contention procedure. The randomized carrier sense interval reduces the chance of collisions on SYNC packet. This process is repeated until all the sensor nodes hear the SYNC packet. Using the broadcast nature of wireless transmissions, the synchronization can be easily performed from the sink to the entire sensor field and each node is required to send the SYNC packet only once. Another function of this SYNC packet is the neighbor discovery service. When a node receives a SYNC packet from its neighbors, it records the address of the sender as one of its neighbors. Since every sensor node sends the SYNC packet at least once, each node can record all of its neighbors. The list of neighbor nodes are used in coordinating the schedule as explained in Section III.2. Since the clock drift on each node can cause synchronization errors, the clock synchronization needs to be performed periodically to avoid synchronization discrepancies. However, the clock synchronization is rarely done since clock drift is very small for typical clock generators used in sensor nodes. Assuming a conservative clock drift rate of 5μs per second and synchronization interval of 1 minute, the maximum clock drift that can be made between a sender and a receiver is about 600μs, which corresponds to the time that can transmit only 1 and half byte assuming 20Kbps transmission bandwidth used in a typical sensor node. This requires a very coarsegrained coordination of time compared to the fine-grained synchronization used in TDMA-based schemes. In cases where a sensor node is not reachable from the sink, a sensor node may not receive the SYNC packet indefinitely. To avoid using up all its energy, the sensor node can be programmed to wait for a fixed amount of time long enough to cover the maximum expected delay that the SYNC packet can arrive with. Eventually the node gives up the synchronization and operates with a lower duty-cycle. However, the node periodically goes back to the active synchronization phase during which the node fully awakes to synchronize its schedule with other nodes. 2) Schedule Synchronization One key issue with AMAC s adaptive cycle time is how a node can synchronize its wakeup period with its neighbors. To accomplish this, the SYNC packet of AMAC also contains the schedule of the sender, i.e. its current cycle time, which is expressed as a multiple of the minimum cycle time T. Initially, this value can be set to the minimum T for fast clock synchronization. Thus, the SYNC packet provides three synchronization features clock, communication, and schedule synchronizations. In addition, every node maintains a schedule table, which records the cycle times of its neighbors. Initially a node can record the cycle time of its neighbor by overhearing the SYNC packet during the initial synchronization phase. Since then, each node updates its schedule table independently every cycle. The absence of the communication SYNC packet suggests that all of its neighbors are idle during the cycle. Thus, each node can keep track of the cycle time adjustment of its neighbor without explicitly asking the schedule information. If a neighbor has been idle and the current cycle is at the neighbor s natural cycle alignment boundary, then the node can update the cycle time of its neighbor by doubling its cycle time. If there is a communication SYNC packet, all the listeners update the schedule of the sender by setting its cycle time to the minimum. In addition, they also update the schedule of the receiver by overhearing the RTS packet. This way, each node can perform the schedule table updates independently every cycle without incurring any extra traffic. The implementation of the adaptive cycle time feature does not require extra overhead except the extra preamble bit and the schedule information added to the SYNC packet. This extra overhead is thus negligible, but as shown in the next section, this simple tuning can dramatically improve both the energy consumption and the performance. IV. EXPERIMENTATION AND RESULTS In this section we evaluate both the latency and the energy performance of AMAC by using detailed packet-level simulations. We chose SMAC as a reference protocol since it is one of the representative sensor network MAC protocols, and AMAC can be viewed as a derivation of SMAC. For this evaluation, the performance optimization features of SMAC such as adaptive listening and message passing are not considered. We rather focus on the impact of the adaptive active period and the adaptive cycle time on the baseline SMAC. Figure 3. The network topology and design parameters used for simulations. A. Simulation Methodology We have implemented the detailed packet-level simulator using ns-2 [11] to model both AMAC and SMAC protocols. The packet formats and protocol parameters used for the simulations are summarized in Fig. 3. Similar metrics and numbers are used in early works [6, 8]. We use a random topology with 100 nodes as shown in Fig.3. The source generates 100 messages, each of which is 200 bytes long. We

vary the traffic load by changing the packet inter-arrival time on the source node from 0 to 10 seconds, where 0 second implies that all the packets are generated once and queued at the same time on the source node. Under each traffic condition, the test is independently carried out 10 times. Each simulation run lasts for an hour. In our simulations SMAC is modeled with a 4T fixed cycle time, which leads to 10% fixed duty cycle, which is in line with SMAC s previous study [8]. In the case of AMAC, three different variations of AMAC are tested: AMAC adaptive active period has the same cycle time as SMAC but with the adaptive active period feature. Its duty cycle is only 3.88% since it removes the RTS/CTS handshaking period. AMAC latency_opt is a full-blown AMAC with T MIN = T and T MAX = 4T. It is called latency optimization version of AMAC since only the cycle time reduction is enabled. Its duty cycle ranges from 3.88% (T MAX = 4T) to 15.5% (T MIN = T). AMAC full is a full blown AMAC with T MIN = T and T MAX from 8T to 64T. Its duty cycle ranges from 0.48% (T MAX = 64T) to 15.5% (T MIN = T). We use two metrics to analyze the performance of AMAC: average per-node dissipated energy and average packet latency. Average per-node dissipated energy measures the total dissipated energy divided by the number of nodes in the sensor field over the simulation period. This metric computes the average energy consumed by a node in delivering 100 messages to the sink. The metric also indicates the overall lifetime of sensor nodes. Average packet latency measures the average communication latency observed for each packet between a source and a sink. The metric indicates the communication performance of the network. B. Communication Latency Fig. 4 shows the average latency of 100 packets as the packet inter-arrival time increases. The inter-arrival time of 0 second implies that all the 100 packets are generated in a burst on the source node. The inter-arrival time of one second means that each packet is generated every second. As the packet interarrival time increases, the latencies of all the MAC schemes start to decrease since the delay due to the contention diminishes. As expected, AMAC latency_opt gives the highest communication performance. Both SMAC and AMAC adaptive active period show the comparable performance and as the packet inter-arrival time reaches 8 seconds, both schemes catch up with the performance of AMAC latency_opt. This demonstrates that the extra overhead required by the adaptive active period feature of AMAC is negligible. In addition, the full-blown AMAC schemes with a larger T MAX show competitive performance for burst traffic patterns with inter-arrival rates of 0 to 6 seconds except the case with the largest cycle time (T MAX = 64T). It turns out that the cycle time reduction feature of AMAC can effectively alleviate the impact of contention in burst traffic patterns since the operation of busy nodes do not collide with the operation of slow nodes. However, when the packet inter-arrival time is long enough, the latency of these full-blown AMAC is higher due to the long hop-to-hop delay incurred by a larger cycle time. Figure 4. The average packet latency in terms of the packet inter-arrival time. C. Energy Consumption Fig. 5 compares the average per-node energy consumption of various versions of AMAC to SMAC in delivering 100 packets. Surprisingly, AMAC adaptive active period can almost halve the energy consumption of SMAC by removing the unnecessary wakeups due to the RTS/CTS listening. Fullblown AMAC schemes can further reduce the energy consumption by increasing the cycle time adaptively. For example, AMAC full with 64T as T MAX consumes only about 14.6% of SMAC s energy consumption. In addition, the pernode energy consumption of these full-blown AMAC schemes start to grow slightly as the packet inter-arrival time increases. This is due to the extra overhead incurred by the longer cycle time adjustment period for sparse traffic scenarios. Figure 5. The average per-node energy consumption in terms of the packet inter-arrival time. Another interesting result is that AMAC latency_opt has slightly lower energy consumption compared to AMAC adaptive active period

even though AMAC latency_opt aggressively reduces the cycle time. It turns out that AMAC latency_opt reduces the cost of overhearing for idle nodes since idle nodes often may not overhear the communications among the busy nodes. This is illustrated in Fig. 6 where the per-node energy consumption of each scheme is break down into costs due to transmissions and receptions, overhearing, and idle listening. Figure 6. Energy distribution of each protocol V. CONCLUSION In this paper we propose a new media access control (MAC) protocol called AMAC that is designed for wireless sensor networks. AMAC is fundamentally different from the existing sensor network MAC protocols in that each node can adjust the duration of its active period as well as the cycle time of the periodic interval depending on the traffic. This leads to a variable duty-cycle operation as compared to the fixed dutycycle operations used in the existing protocols. This variable duty-cycle operation allows us to achieve both high performance and low energy consumption at the same time since busy nodes can work with the highest duty-cycle while idle nodes can work with the lowest duty-cycle. The simulation results confirm that the features of AMAC are very effective in reducing the energy consumption as well as in improving the network performance of burst traffic patterns. REFERENCES [1] Ding, J., Sivalingam, K. M., Kashyapa, R., and Chuan, L. J., "A multilayered architecture and protocols for large-scale wireless sensor networks," in Proc. IEEE Semiannual Vehicular Technology Conference-Fall, Oct. 2003 [2] Heinzelman, W. R., Chandrakasan, A., Balakrishnan, H., Energy- Efficient Communication Protocol for Wireless Microsensor Networks, in Proceedings of the 33rd Hawaii International Conference on System Sciences, Vol. 2, pp. 10, 2000 [3] Sohrabi, K., Gao, J., Ailawadhi, V., and Pottie, G., A Self Organizing Wireless Sensor Network, in Proceedings of the 37th Allerton Conference on Communication, Control, and Computing, Sept. 1999 [4] Christian C. Enz., Amre, E. H., Decotignie, J. D., Peiris, V., WiseNET: An Ultra Low-Power Wireless Sensor Network Solution, IEEE computer Magazine, Vol. 37, No. 8, Aug 2004 [5] Dam, T., Langendoen, K., An adaptive energy-efficient MAC protocol for wireless sensor networks, in Proceedings of the 1st International Conference on Embedded Networked Sensor Systems, 2003 [6] Ye, W., Heidemann, J., Estrin, D., Medium Access Control with Coordinated Adaptive Sleeping for Wireless Sensor Networks, IEEE Transactions on Networking, Vol. 12, pp. 493-506, Issue. 3, 2004 [7] Lin. P., Qiao. C., Wang. X., Medium access control with a dynamic duty cycle for sensor networks, in Proceedings of the IEEE Wireless Communications and Networking Conference, March 2004 [8] Ye, W., Heidemann, J., Estrin, D., An Energy-Efficient MAC Protocol for Wireless Sensor Network, in Proceedings of the IEEE Infocom, pp.1567-1576, June 2002 [9] A. Woo and D. Culler, A Transmission Control Scheme for Media Access in Sensor Networks, in Proceedings of the 7th Annual International Conference on Mobile Computing and networking, 2001 [10] TinyOS Project http://www.tinyos.net [11] The Network Simulator ns-2, http://www.isi.edu/nsman/ns