MAC Protocols for mobile wireless sensor networks

1. MAC Protocols for mobile wireless sensor networks • Luís Bernardo • Miguel Pereira • Francisco Ganhão • Rodolfo Oliveira • Rui Dinis • Paulo Pinto July 5, 2010 Ciência 2010 tele1.dee.fct.unl.pt

2. Motivation MAC layer PHY layer Conclusions Outline

3. Motivation • Critical infrastructure protection with wireless sensor networks • Contradictory objectives: • Maximize WSN lifetime (minimize energy consumption) • Have controlled packet delay and throughput • Support mobile and fixed battery powered nodes

4. Motivation • Layered approach: 6lowPAN, ROLL, 802.15.4, … • Cross-layer interfaces to handle hardware/energy limitations • MAC layer (Multimode MAC protocols) • Adapt operation to application requirements • PHY layer (PC H-ARQ / MPD receivers) • Reduce energy lost with collisions/interference satisfying app. requirements Objective: Save Energy and meet the application’s communication requirements e.g. Vehicle tracking vs. Environment monitoring Application Transport Routing MAC PHY

5. MAC - Motivation A Wireless Sensor Network (WSN) mobility scenario Mobile nodes moving through static WSN islands Static nodes (single radio) - battery must be saved Mobile nodes - external energy resources High throughput needed during a short connection Standard WSN Medium Access Control (MAC) protocol do not handle the set of requirements mentioned before

6. MAC - Motivation • Medium Access Control (MAC) protocols save Energy by turning the radio off • Asynchronous MAC protocols (e.g. B-MAC; X-MAC) • Low Power Listening bind the receiver and sender using a large preamble • Advantages • Nodes run independent asynchronous duty-cyles - good for mobility • Energy efficient to bursty traffic • Disadvantages • Limited throughput and high delay for more than one sender

7. MAC - Motivation • Synchronous MAC protocols (e.g. S-MAC, LL-MAC, Z-MAC, 802.15.4) • Contention protocols using Carrier Sense Multiple Access (e.g. S-MAC) • Scheduled protocols using Time Division Multiple Access (e.g. LL-MAC) • Use hybrid approach (e.g. Z-MAC) • Support both: CSMA and TDMA - changes to TDMA fallback during load peaks, maximizing the throughput • Advantages • High throughput available for peak periods

8. MAC - Motivation • Disadvantages • High energy consumption even for idle periods • Synchronized duty-cyles - bad for mobility • CSMA requires SYNC frame before communication • TDMA requires an additional slot allocation algorithm • High mobility requires high SYNC rates to keep track from the neighbors

9. MAC - Conceptual Idea • Goal • Have a low energy asynchronous mode • Have a synchronous mode high throughput in the presence of mobile asynchronous nodes • Allow shorter connection times than other hybrid protocols • Maximize throughput for mobile nodes in the neighborhood of synchronous nodes • We propose the Mobile Multimode Hybrid MAC (MMH-MAC) • Asynchronous and Synchronous modes

10. Asynchronous Mode • Goal • Minimize the idle energy consumption • MMH-MAC asynchronous mode uses • Preamble sampling approach similar X-MAC protocol • Two techniques to minimize the interference between synchronous and asynchronous nodes • It uses Low Power Listening mechanism • Sender sends a sequence of short preambles with duration up to 2*Tduty_cycle before the data frames • Unicast receivers may send and Early Preamble ACK

11. Asynchronous Mode • Passive interference mitigation • Alignment of the asynchronous active time with the public slot of the last visited synchronous node • Preamble overhead is reduced due to the immediate reception of an early PACK • Active interference mitigation • Improved Shut-up mechanism

12. Synchronous Mode • Slotted scheme - Nodes runs a synchronized duty-cycle period. • 11 slots with fixed duration of 100ms each • Slots are subdivided in ten 10ms subslots • Public Slot (slot 0) • Shared by all the nodes, it’s used for broadcast traffic and casual unicast traffic • Unicast traffic is acknowledged and run a contention based protocol • First 50 ms reserved for MAC signaling (SYNC frames)

13. Synchronous Mode • Private slot (slot 1-10) • Reserved slots for unicast traffic between two nodes • Collision free environment • Traffic is acknowledged • After 25ms of inactivity nodes go into sleep • SYNC frames are used to: • Maintain inter-node duty-cycle synchronization • Broadcast private slot allocation • As beacons to detect neighborhood changes (above an RSSI value)

14. Synchronous Mode • MMH-MAC mobility handling features • Multiple SYNC frames can be transmitted per duty-cycle • Normal SYNC frames are transmitted in a random subslot of public slot 0 • Other SYNC frames are sent when an asynchronous node is detected • A neighbor SYNC table is kept that measures link stability allowing cluster formations

15. Synchronization Process • Goal • Guarantee that all neighbors follow the same duty-cycle schedule (synchronous and asynchronous nodes) • If all nodes are asynchronous • Packet Hello is sent preceded by a sequence of preambles • Request/Ok exchange identifies the neighbors and reserves private slots • SYNC defines the initial synchronization reference

16. Synchronization Process • If at least one node is synchronous, neighbor nodes follow the existing duty cycle • Passive approach (classical) • Where M node waits for the SYNC packet • Active approach (new) • M sends preambles to trigger the Shup-Up mechanism in one active slot in one of its neighbors • Wait for the SYNC to proceed with the synchronization • First empty slot or idle dedicated slot • Next public slot

17. Synchronization Process • Performance • Depends on the number of active private slots • more active slots = less time a node takes to listen to M preambles • more active slots = more time until finding an idle slot • MMH-MAC proposes the use of listening private slots mechanism • The node turns on the radio for 10 ms when the slot is free • Each listening slot costs 1% of duty-cycle • Depends on the preamble starting slot • Slot 0 is the optimal case

18. MAC - Results • We use TOSSIM simulator • Run MMH-MAC nesC code • Added the mobility support • Additional meters measure active time/sleep time/tx time/receive time • Simulated scenario • 21 static nodes in synchronous mode organized in 6 static clusters • Each dedicated slot has CBR traffic (10 packets/sec and 35 bytes/packet) • Each static node sends one SYNC per duty-cycle (1,1s minimum value) • Energy estimation: Xbow Telos B current consumption

19. MAC - Results • Simulated scenario • A mobile node moves randomly on the scattered WSN • Connects 120 times to the islands with a variable connection time • We evaluate three scenarios [WCNC’2010] • Passive syncronization • Active synchronization without listening slots • Active synchronization with one listening slot (slot 6)

20. MAC - Results • Time to synchronize • As function of the number of allocated dedicated slots

21. MAC - Results • Throughput • As function of the connection duration time

22. MAC - Conclusions • MMH-MAC significantly reduces the time to an asynchronous node to start communicating to a synchronous node and vice versa • Minimize the interference between asynchronous and synchronous nodes • We implement the code on TinyOS and we made short tests on real nodes • We are implementing a mixed TelosB / SunSPOT scenario

23. PHY - Motivation • Classical WSN PHY (e.g. 802.15.4) limit energy efficiency • Packets involved in collisions/interference are lost • Low complexity H-ARQ may improve energy efficiency • WSN applications with hard constraints on: • Delay • Bitrate

24. PHY - Motivation • Using an H-ARQ scheme enhances the throughput, compared to a conventional ARQ scheme; • Energy could be saved on subsequent re-transmissions; • Depending on the distance and the nodes density: • Circuit’s energy consumption ≥ expended energy transmission.

25. PHY - Objectives • Analyze the Energy per useful packet (EPUP): • Diversity Combining (DC) H-ARQ technique; • Conventional ARQ (C-ARQ); • Obtain the optimal EPUP for a TDMA access mode considering: • Delay constraints • Throughput constraints.

26. PHY - System Overview • Assumptions: • Synchronous TDMA MAC slot on a flat fading scenario; • Additive White Gaussian Noise channel (AWGN); • Slots of equal length, each equivalent to a packet of M bits; • A receiver, holds up to R transmissions of a failed packet; • After R transmissions, it gives up.

27. PHY - System Overview • Receiver Characterization for DC H-ARQ: • Linear Bit Combination; • Enhancement of the bit reception. . .. +

28. PHY - System Overview • Energy Analysis EPUP – Energy per useful packet E[N] – Expected number of retransmissions Ep – Energy per Packet(d, Eb) QR+1 – Probability of packet failure after R transmissions

29. PHY - System Overview • System Optimization - minimize EPUP, subject to: • A minimum goodput Smin ; • A maximum delay Dmax ; • A minimum success probability.

30. PHY - Performance • C-ARQ vs. DC H-ARQ [ICCCN’2010a]: • Analytical and simulated results with the ns-2 simulator; • Simulation characteristics: • Packet size of M=256 bits; • 8 Wireless Terminals; • Distances ranging between d=10m and 100m; • Retransmissions up to R=10.

31. PHY - Performance • EPUP in function of d and Eb/N0.

32. PHY - Performance Success Probability Delay

33. PHY - Conclusions • DC H-ARQ can extend the battery of a Wireless Terminal, compared to a conventional TDMA ARQ scheme. • Longer distances; • Re-transmission tolerance. • Future Work: • MultiPacket Detection schemes [Globecom’07, TWC09, ICCCN’2010b]

34. PHY – MPD vs DC H-ARQ Delay Throughput

35. Conclusions & Future Work • MAC layer approaches adapt radio sleep times and synchronization to the application/routing requirements • PHY layer reduce transmission power, or synchronization requirements, by using DC H-ARQ or MPD • Future Work: • Continue to combine MAC and PHY approaches to improve energy efficiency

36. Thank you for your attentionQ & A