1. A+MAC: A Streamlined Variable Duty-Cycle MAC Protocol for Wireless Sensor Networks 1 Sang Hoon Lee, 2 Byung Joon Park and 1 Lynn Choi 1 School of Electrical Engineering, Korea University {smile97, lchoi}@korea.ac.kr 2 Department of Computer Science, Kwangwoon University bjpark@kw.ac.kr In Proc. Computer and Applications (CCA) 2012 To appear International Journal of Distributed Sensor Networks (SCIE)

2. Motivation Two motivations Control packets contain unnecessary information. Active period is still long. Ideas Simplification of CSMA/CA We eliminate unnecessary information from control packets. Combined SYNC/RTS Period Wakeup preamble period A short wakeup period is designated to announce the presence of traffic.

3. Idea 1: Simplification of CSMA/CA Remove the unnecessary information from control packets No virtual carrier sensing Nodes, which don’t participate in a communication, sleep until the next cycle time We need no duration field in control packets. No transmitter address field in CTS and ACK packets The destination of CTS and ACK already knows who will send the control packet. SMAC versus A+MAC: control packet formats

4. Idea 2: Combined SYNC/RTS Period Reduce the length of an active period Each node seldom transmits a SYNC packet. The contention between a SYNC packet and a RTS packet may be very light. Nodes overuse energy to prevent the contention. SMAC versus A+MAC

5. Idea 3: Wakeup Preamble Period Remove the unnecessary long idle listening during idle cycle Exploit a complementary cooperation between preamble sampling and CSMA/CA Contention resolution ability and synchronized schedules of CSMA/CA based MAC Short idle listening of preamble sampling Wakeup preamble A sender transmits it to notify only the presence of traffic. Contains no information. Therefore, A+MAC tolerates collisions between multiple wakeup preambles. If a node receives a wakeup preamble or detects a collision during a wakeup preamble period, it extends its active period. Since each node coordinates wakeup schedule with neighbor nodes, a sender can send a short preamble just at the receiver’s wakeup preamble period.

6. A+MAC vs. AMAC A+MAC AMAC

7. S-MAC vs. AMAC vs. A+MAC S-MACAMACA+MAC Control packetsAll packets are 9B.Same to S-MACSimplified format RTS: 7B CTS, ACK: 5B SYNC: 9B Wakeup preamble: 1B Notifying the presence of traffic RTSCommunication SYNCWakeup preamble The length of idle listening Separate SYNC/RTS (64ms in NS2) SYNC period (32ms in NS2) Slot time (1ms in NS2) Receiver’s active period Separate SYNC/RTS (64ms in NS2) Separate SYNC/RTS (64ms in NS2) Combined SYNC/RTS (32ms in NS2) Dynamic duty cyclingNoneAdaptive Active Period Adaptive Cycle Time Wakeup Preamble Period Adaptive Cycle Time

8. Experimentation The network topology and design parameters used for NS-2 simulations Simulated protocols S-MAC A-MAC Our previous study for an energy efficient MAC protocol Dynamic duty cycling A+MAC ParameterValue Number of nodes200 random nodes Number of messages100 messages per a source Message size100 bytes Packet generation interval0 to 10 seconds Duty cycle of S-MAC11% (4T) Maximum duty cycle of A+MAC0.04% to 0.31% (32T to 4T)

9. Average Packet Latency The average latency of 100 packets The longer the packet inter-arrival time, the shorter the latency The delay due to the contention diminishes. The cycle time reduction can effectively alleviate the impact of contention in burst traffic patterns. The operation of busy nodes do not collide with the operation of slow nodes. When the packet inter-arrival time is long enough The latency of A+MAC becomes higher due to the long hop delay incurred by a larger maximum cycle time.

10. Energy Consumption The average per-node energy consumption in delivering 100 packets AMAC can almost halve the energy consumption of S-MAC. AMAC removes the unnecessary wakeups due to the RTS/CTS. A+MAC more aggressively reduces the energy consumption due to idle listening. By eliminating idle listening due to SYNC/RTS period from active period, A+MAC 4T can reduce the energy by more than 98% compared to SMAC although it can also substantially reduce the packet latency.

11. Empirical Evaluation The network topology and design parameters used for MICA2 motes simulations Simulated protocols S-MAC A+MAC ParameterValue Sensor nodesMICA2 motes with CC1000 transceiver Topology14 linear nodes A sink and a source are located at both ends. Number of messages20 messages Message size100 bytes Packet generation interval1 to 10 seconds Duty cycle of S-MAC11% (4T) Maximum duty cycle of A+MAC0.04% to 0.31% (32T to 4T)

12. Average Packet Latency The average latency of 20 packets A+MAC 4T has the highest communication performance. It enhances the performance up to 13.5% from SMAC since the adaptive cycle time can decrease the impact of sleep delay on the message latency. As the packet inter-arrival time increases, SMAC catches up A+MAC 8T. In other words, A+MAC 8T can reduce the number of nodes’ wakeup while it provides comparable communication performance with S-MAC.

13. Energy Consumption The average per-node energy consumption in delivering 20 packets We simulated the energy consumption of a node with AEON tool. The energy model of AEON is based on measurements of node current draw and the execution of real code. A+MAC 4T can decrease 90% of energy consumption of SMAC. The diminution in the energy consumption of A+MAC is smaller than that in the result of NS-2 simulation. The impact of overhearing on the energy consumption is not observed

14. Conclusion A+MAC is fundamentally different from the existing MAC protocols A complementary cooperation between preamble sampling and CSMA/CA Each node can dynamically adjust the operation of MAC protocol. Duration of an active period Duration of a periodic interval (cycle time) Variable duty-cycle operation with wakeup preamble allows us to achieve both high performance and low energy consumption at the same time. Busy nodes can work with the highest duty-cycle. Idle nodes can work with the lowest duty-cycle and minimize the idle listening.

15. Q&A