1. A Transmission Control Scheme for Media Access in Sensor Networks Presented by Jianhua Shao

2. Overview  Application and platform characteristics of sensor network and corresponding challenges  Hardware and software platform of design  Related work on existing MAC and shortcomings for sensor network  Proposed MAC and transmission control scheme  Simulation in single hop  Simulation in multihop

3. Traffic Characteristics in Sensor Network  Sample environment for sensor information  Propagate data back to infrastructure  Little traffic over long periods while very intense for short periods  Periodic sampling yields high amount of highly correlated traffic  Originating traffic less than route-thru traffic  Collective structure and multihop topology

4. Related Work  Many CSMA strategies not enough for sensor networks  Packet transmissions occur with a random distribution, rather than correlated traffic  Support independent point-to-point flows, rather than collective structure

5. Related Work  802.22  A wireless Ethernet illusion  A single cell scenario but no multihop scenario  Peer-to-peer communication rather than many-to-one data propagation  Primary mechanism of carrier sensing and contention control scheme

6. Related Work  Bluetooth  Emerging standard for wireless devices  A centralized Time Division Multiple Access (TDMA) protocol  Tight requirement of time synchronization  No multihop scenario

7. Related Work  MACAW  Share a single channel radio similar to sensor networks  Single hop base station interaction within a cell  Hidden node problem in a multihop network  Primitive contention control protocol

8. How to evaluate -- Metrics  High channel utilization  Efficient energy consumption  Fair allocation of bandwidth

9. Sensor Network in TinyOS

10. Platform of Sensor Network  ATMEL 4MHz 8 bit microprocessor  8K program memory  512 bytes data memory  Single channel RF transceiver  Operating at 916MHz  10kbps using on-off-keying encoding  Variety of Sensors  Temperature, photo, etc.  TinyOS – event-based operating system  30 byte messages

11. Listening Mechanism in Design  CSMA  All nodes can hear each other, no hidden  Shorten length of carrier sensing – turn off  Collision Detection (CD)  Impossible in wireless network without additional circuits  Highly synchronization leads to collision  Random delay for transmission to unsynchronize the nodes

12. Backoff Mechanism in Design  To reduce contention  Restrain a node from accessing the channel for a period of time  Channel is free after backoff period  A phase shift to the periodicity of application

13. Contention Based Mechanism in Design  Minimizing number of control packets  Eliminate ACK  Only RTS and CTS  Perform handshakes  Send RTS and wait for CTS  If no CTS, backoff  If CTS before transmission, defer

14. Rate Control Mechanism in Design  Balance between originating traffic and route- thru traffic  MAC control rate of originating data of a node  MAC control rate of route-thru traffic  Originating data adaptive to route-thru traffic  Successful injection then increasing rate  Unsuccessful injection then decreasing rate  Route-thru traffic adaptive to originating data  Lots of data injection then decreasing

15. A linear increase and multiplicative decrease approach  To control application transmission rate  Given  Transmission rate S  Probability of transmission p, p  [0, 1]  Originating data rate S*p  Constant , competition degree for channel  Factor , 0 <  < 1, penalty for transmission failure

16. A linear increase and multiplicative decrease approach  How to work  To linearly increase the rate Increment p by a   To multiplicatively decrease the rate Multiply p by a   Choice of  and   originate =  route / (n+1) where n children  route =  originate * 1.5

17. Simulation – Single Hop  Different CSMA Scheme

18. Simulation – Single Hop  A single cell topology for evaluating CSMA

19. Simulation – Single Hop  Settings  Packet size: 30 bytes  Channel capacity: 10kbps  Deliver at most 20.8 packet/s  16-bit CRC error check  Highly synchronized traffic – all nodes start at the same time

20. Simulation – Single Hop  Delivered Bandwidth under Simulation  Each ode attempting to send periodically at 5 packets/s  All CSMA schemes achieve greater bandwidth than the 802.11 scheme  Randomness along with collision detection hardware

21. Simulation – Single Hop  Results for simple CSMA scheme  Good channel utilization  High offered load  Insensitive to backoff mechanism  Randomness in the pre-collision phase is essential for robustness

22. Simulation – Single Hop  Energy usage  Separate portion of energy consumption In transmitting and receiving packets, which is determined by the traffic load In listening, which is determined by the CSMA protocol  Average energy per packet in listening 802.11 has the worst energy efficiency CSMA with constant listen period are most efficient, 10uJ/packet CSMA with random listen period are more costly, at 40uJ/packet

23. Simulation – Single Hop  Energy Usage  Delay no energy since radio off  Most energy efficient schemes are those with constant listen period and a random delay

24. Simulation – Single Hop  Fairness  Fairness at uniform load Differences in backoff is insignificant 802.11 unfair allocation of bandwidth  Proportional fairness Backoff mechanism has an effect Binary exponential and 802.11 worst

25. Simulation – Single Hop  Sensor Phase Shifting  CSMA vulnerable to the capturing effect The transmission fail back to application  CSMA includes an application level adaptation If transmission failure, the phase of the sensor sampling interval is shifted by a random amount Break away from unfortunate synchrony

26. Simulation – Single Hop  Sensor Phase Shifting  Incorporate the phase-shift in to 802.11 Bandwidth and fairness improve substantially

27. Simulation – Single Hop  Empirical Results  Compare the three CSMA schemes with random delay to the simulation result  With each node sending at a rate of 5 packets/s Empirical measurement closely matches the simulation prediction: 70% Average energy spent Fairness

28. Simulation – Single Hop  Empirical Results  When correlated nodes transmit at the same time and if no randomness, no successful transmission is possible

29. Simulation – Single Hop  CSMA Scheme conclusion  Random delay  A constant listen period  Radio powered down during backoff  Phase-shifting if half-duplex network stack  Backoff good for proportional fairness, not good for aggregate bandwidth and fairness  Suggestion for backoff

30. Simulation – Multihop  Multihop topology

31. Simulation - Multihop  Two challenges  Too much traffic for near nodes, little bandwidth for distant nodes  Too much traffic for distant nodes, packets dropped and routing wasted

32. Simulation - Multihop  Four considered schemes  CSMA augmented with a transmission control protocol: D_CONST_FIX Fair share to downstream and upstream  A traditional RTS/CTS contention control scheme  An adaptive rate control (ARC) scheme  802.11 CSMA with ACK

33. Simulation – Multihop  Assumptions for running the simulation  each node sending packets to the base station at rate of 4 packets/s with the same start time  The base station will echo each packet it receives in all schemes to make a fair comparison

34. Simulation - Multihop  Bandwidth delivered to the base station from each node  The two basic CSMA schemes fail to deliver any packets from nodes which are more then two level deep  RTS/CTS can deliver in such situation, but unfair bandwidth allocation  ARC provides the most fair delivered bandwidth

35. Simulation - Multihop

36.  Different  and  in controlling the tradeoff among fairness, energy efficiency and aggregate bandwidth  Always lower variance of fairness for the adaptive scheme   increase while  irrelevant  When  < 0.2,  is more important A small  leads to a conservative scheme A large  impose a smaller penalty

37. Simulation - Multihop  Fairness

38. Simulation - Multihop  Aggregate bandwidth

39. Simulation - Multihop  Energy efficiency

40. Conclusion  Adaptive rate control can effectively achieve fairness of bandwidth allocation  Energy efficient – no control packets