Presentation is loading. Please wait.

Presentation is loading. Please wait.

Fakultät Informatik – Institut für Systemarchitektur – Professur Rechnernetze Sensor Medium Access Control (S-MAC) Robin Dunn Supervisor: Dr. Waltenegus.

Similar presentations


Presentation on theme: "Fakultät Informatik – Institut für Systemarchitektur – Professur Rechnernetze Sensor Medium Access Control (S-MAC) Robin Dunn Supervisor: Dr. Waltenegus."— Presentation transcript:

1 Fakultät Informatik – Institut für Systemarchitektur – Professur Rechnernetze Sensor Medium Access Control (S-MAC) Robin Dunn Supervisor: Dr. Waltenegus Dargie

2 9 Jan. 06 Sensor MAC 2 Of 32 Outline Motivation Protocol design – Periodic Listen & Sleep – Adaptive Listening – Overhearing Avoidance – Message Passing Protocol Implementations Experimentation Summary References

3 9 Jan. 06 Sensor MAC 3 Of 32 Motivation In a WSN: “energy conservation and self-configuration are primary goals...” [1]

4 9 Jan. 06 Sensor MAC 4 Of 32 Motivation Need to reduce energy waste from: Collision Overhearing Control overhead Idle Listening Need to have a robust topology to: Account for lost nodes New joining nodes

5 9 Jan. 06 Sensor MAC 5 Of 32 Periodic Listen and Sleep(1) Choosing & Maintaining Schedules Nodes choose their own schedules or follow others Neighbouring nodes synchronise – virtual clusters – Reduces latency – SYNC packet to communicate schedules Address of sender Time to next sleep – Duration of sleep not transmitted – user defined at compile time – Schedules of neighbours stored in schedule table [2]

6 9 Jan. 06 Sensor MAC 6 Of 32 Choosing & Maintaining schedules Boot up Node 'A' Listens Receives no SYNC so chooses and follows own schedule When a node has chosen or adopted a schedule it broadcasts its schedule. Boot up SYNC period ASYNCCS BSYNC Receives SYNC from node B so follows B's schedule ASYNC CS SYNC period t t x SYNC period >= x

7 9 Jan. 06 Sensor MAC 7 Of 32 Periodic Listen and Sleep(2) Choosing & Maintaining Schedules – What if a node receives a different schedule after choosing own? A SYNC Receives no SYNC from another node so chooses and follows own schedule B SYNC But then receives SYNC from another node B. CS ASYNC BSYNC B If A has no other neighbours: A 1. A listens...hears no SYNC... 2. A Chooses own schedule... 3. Broadcasts SYNC... 4. B joins... 5. B broadcasts own SYNC... 6. A adopts B's schedule. Node 'A' Listens t

8 9 Jan. 06 Sensor MAC 8 Of 32 Periodic Listen and Sleep(2) Choosing & Maintaining Schedules – What if a node receives a different schedule after choosing own? If A has at least one neighbour: ASYNC BSYNC B NANA NANA A 1. A listens...hears no SYNC... 2. A Chooses own schedule... 3. Broadcasts SYNC... 4. A following schedule with neighbours... 5. B joins... 6. B broadcasts own SYNC... 7. A adopts both schedules. A SYNC Receives no SYNC from another node so chooses and follows own schedule B SYNC CS Node 'A' Listens t But then receives SYNC from another node B. t

9 9 Jan. 06 Sensor MAC 9 Of 32 Periodic Listen and Sleep(3) Choosing & Maintaining Schedules Boarder nodes adopt both schedules. One neighbour so only follows one schedule. Form virtual clusters but communicate with peers No Cluster head. Border nodes have less time to sleep. Periodic neighbour discovery nodes periodically listen for the whole SYNC period [2]

10 9 Jan. 06 Sensor MAC 10 Of 32 Periodic Listen and Sleep(4) Maintaining Synchronization Clock drift → synchronisation errors Avoid these errors using – Relative Timestamps – time to next sleep – Listen times longer than clock drift rate – Periodic synchronisation updates. SYNC packet (same as on Boot Up). – Address of sender – Time of senders next sleep Node A listen Node B listen

11 9 Jan. 06 Sensor MAC 11 Of 32 Periodic Listen and Sleep(5) Timing for sending and Receiving data Contention – Node B wants to send data to node A Listen period Carrier Sense is for random duration Node A Node B Transmit data if no current transmission t t CSRTS CTS DATA ACK CS = Carrier Sense RTS = Request to send CTS= Clear to send

12 9 Jan. 06 Sensor MAC 12 Of 32 Periodic Listen and Sleep(6) Timing for sending and Receiving data Division of the listen interval – To ensure both SYNC and Data packets are received [2] Receive Data

13 9 Jan. 06 Sensor MAC 13 Of 32 Adaptive Listening - Problem Delay caused by strict adherence to sleep schedules A B C Event Sensed B wants to pass the message on to C. It knows when C's next listen time is but it doesn't want to wait until then. It wants to send the message immediately! B waiting for C to wake up... C wakes up...

14 9 Jan. 06 Sensor MAC 14 Of 32 Adaptive Listening - Solution Adaptive Listen mechanism to reduce delay Duration field included in control packets (RTS/CTS/ACK) A B C Event Sensed Network Allocation vector (NAV) – count down to end of transmission As soon as B has received all the data from A, C will wake up immediately to receive the data from B Receives RTS from B and so on... If C missed RTS/CTS it can get duration from ACKs CTS C now knows from duration field when the transmission will end C wakes up at end of transmission

15 9 Jan. 06 Sensor MAC 15 Of 32 Adaptive Listening Adaptive Listen mechanism to reduce delay - summary RTS Node A CTS DATA AL RTS = Request to send CTS= Clear to send AL = Adaptive Listen Node B Node C ABC RTS CTS DATA AL RTS And so on... Listen t t t t Without AL A could be sleeping at this moment

16 9 Jan. 06 Sensor MAC 16 Of 32 Overhearing Avoidance Problem: A node listens to data transmissions even though it knows it is not involved. Solution: Nodes sleep after RTS, CTS or ACK Which nodes sleep? B sleeps because: A→B collides with C→B (RTS,DATA) B→A collides with D→C (ACK) E sleeps because: E→D collides with C→D (RTS,/DATA) All immediate neighbours of transmitting node sleep Nodes know when to wake up by using NAV ABCDEF x x [2]

17 9 Jan. 06 Sensor MAC 17 Of 32 Message Passing(1) Efficient transmission of long messages. Other Approaches: – One long packet. Corruption of part means resending all. – Many small packets with RTS,CTS,DATA,ACK for each. Large control overhead and delay though contention for each.

18 9 Jan. 06 Sensor MAC 18 Of 32 Message Passing(2) Message Passing Approach: Fragmented message transmitted as burst 1RTS. 1CTS. Multiple ACKs. RTS CTS DATA ACK DATA ACK Sending Node Receiving Node Extension of transmission time for resending of corrupted packets Limit on number of extensions to account for dead receiver. Multiple ACKs required for AL and overhearing avoidance

19 9 Jan. 06 Sensor MAC 19 Of 32 Message Passing (3) Advantages – Smaller delay when packets are corrupted i.e. Can resend just corrupted packets – Smaller control overhead – Favours message fairness over node fairness – Sending multiple ACKs prevents hidden terminal problem: A BC C cant here A's transmission C thinks B is free A B C ACK DATA C hears ACK Realises B is busy Sets NAV

20 9 Jan. 06 Sensor MAC 20 Of 32 Implementation 2 Implementations carried out: – 1 st Rene Motes (No Adaptive Listening). 2 hop network. [1] – 2 nd Mica Motes (Adaptive Listening Option) 10 hop network. [2] – Both using TinyOS User can configure options when compiling: – Duty Cycle selection (1%-99%) – Fully Active Mode (periodic sleep disabled) – Disable adaptive listen Current Implementations coordinate radio sleeping. Further work required to put other hardware to sleep (e.g. CPU). – Further energy savings

21 9 Jan. 06 Sensor MAC 21 Of 32 Experimentation(1) [2] Source 1 Sink 2 Sink 1 [2] Biggest energy savings when low traffic 1 st Implementation. Rene Motes 2 hop network. Mean energy consumption on radios in each source node

22 9 Jan. 06 Sensor MAC 22 Of 32 Experimentation(2) 2 nd Implementation Mica Motes 10 hop network [2] Energy consumption results including adaptive listen Again biggest energy savings during low traffic [2]

23 9 Jan. 06 Sensor MAC 23 Of 32 Experimentation (3) Latency - measured from the time a message is generated on the source node [2] Mean message latency on each hop - Low traffic Mean message latency on each hop - High traffic AL – Reduces latency [2]

24 9 Jan. 06 Sensor MAC 24 Of 32 Experimentation(4) Throughput – only data packets counted, not control packets. Throughput at maximum traffic load AL improves throughput [2]

25 9 Jan. 06 Sensor MAC 25 Of 32 Experimentation(5) Throughput at different traffic loads AL improves throughput Throughput can not be increased further because not enough traffic Throughput – only data packets counted, not control packets. [2]

26 9 Jan. 06 Sensor MAC 26 Of 32 Experimentation (6) Combined affect of energy consumption and reduced throughput The benefits of adaptive listen occur at moderate to high traffic load. [2]

27 9 Jan. 06 Sensor MAC 27 Of 32 Summary SMAC makes energy savings by periodically putting nodes to sleep. This can result in higher latency and reduced throughput But these problems are mitigated by the use of Adaptive Listening

28 9 Jan. 06 Sensor MAC 28 Of 32 Summary Motivation Protocol design Periodic Listen & Sleep Adaptive Listening Overhearing Avoidance Message Passing Implementations Experimentation Summary References

29 9 Jan. 06 Sensor MAC 29 Of 32 References(1) Main references: [1] An Energy-Efficient MAC Protocol for WirelessSensor Networks. Wei Ye, John Heidemann, Deborah Estrin. In Proceedings of the 21st International Annual Joint Conference of the IEEE Computer and Communications Societies (INFOCOM), Vol.3, pp. 1567-1576, New York, NY, USA, June, 2002. [2] Medium Access Control With Coordinated Adaptive Sleeping for Wireless Sensor Networks. Wei Ye, John Heidemann and Deborah Estrin IEEE/ACM Transactions on Networking, Vol. 12, No. 3, pp. 493-506, June 2004. SMAC Source code available to download at http://www.isi.edu/ilense/software/smac/download.html http://www.isi.edu/ilense/software/smac/download.html

30 9 Jan. 06 Sensor MAC 30 Of 32 References(2) Related Work: Energy and Latency Control in Low Duty Cycle MAC Protocols. Yuan Li, Wei Ye, John Heidemann. IEEE Wireless Communications and Networking Conference (WCNC), March 2005, New Orleans, LA. Implementation of the Sensor-MAC protocol for the JiST/SWANS simulator. Dissertation. Tippanagoudar, Veerendra. FLORIDA ATLANTIC UNIVERSITY. 2006. http://wwwlib.umi.com/dissertations/preview/1436493 Have to pay to get full access!http://wwwlib.umi.com/dissertations/preview/1436493 SMAC Implementation for ETRI SSN-based Sensor Networks. Kiran Tatapudi and Chansu Yu Department of Electrical and Computer Engineering Cleveland State University. http://www.csuohio.edu/ece/techReport/2006/tr03.pdf http://www.csuohio.edu/ece/techReport/2006/tr03.pdf Adaptive Forwarding Error Correction Code Control: An Adaptive FEC Code Control Algorithm for Mobile Wireless Sensor Networks. JOURNAL OF COMMUNICATIONS AND NETWORKS, VOL. 7, NO. 4, DECEMBER 2005. http://network.dongguk.ac.kr/ http://network.dongguk.ac.kr/


Download ppt "Fakultät Informatik – Institut für Systemarchitektur – Professur Rechnernetze Sensor Medium Access Control (S-MAC) Robin Dunn Supervisor: Dr. Waltenegus."

Similar presentations


Ads by Google