MERLIN: A synergetic Integration of MAC and Routing for Distributed Sensor Networks

1. MERLIN: A synergetic Integration of MAC and Routing for Distributed Sensor Networks A.G.Ruzzelli, M.J.O’Grady, R.Tynan, G.M.P.O’Hare. Adaptive Information Cluster project (AIC) Smart Media Institute (SMI) Department of Computer Science University College Dublin Ireland. http://www.adaptiveinformation.ie/home.asp

2. Summary • Overview of WNSs and protocols • Phase1: MERLIN design • Motivation and objectives • Fundamental concept • MAC details • Routing details • Phase2: Simulation and results • Scheduling performance • Comparison against SMAC+ESR Conclusion

3. Overview of Wireless Sensor Networks • Large number of tiny sensors (nodes) distributed in an area network; • Sensor nodes: • have sensing devices attached; • are self-organizing; • are usually battery operated and of low cost hence power limited  • multi-hop communication to save energy;

4. WSNs issues: • Nodes must be cheap  Limited memory capabilities •  Limited processing capabilities •  Limited power capabilities •  Maybe not very reliable • Limited energy  Power consumption • High node number  Scalability issues • High dynamic condition Reactivity and Self-organization • Always on radio  node depletionin few days (e.g. Mobile phone) • MAC issues • Simultaneous msg to same device Packet Collision • Channel access delay • Control packet overhead • Multihop routing issues • Route maintenance Overhead • E-to-E latency • Global addressing issue • Only useful messages to deliverIn-network processing

5. Sensing devices Application Data aggregation Sensing coverage Localization Cross layer interaction Routing MAC Physical Antenna Wireless sensor network architectureAn example: • The most suitable network architecture for WSNs is still an open issue • Each layer has its own task • Any layer try to achieve the task using the smallest amount of energy possible • Researchers are evaluating how to use the cross layer interaction at best

6. Mechanisms applied in Wireless Sensor Networks MACs: • The CSMA/CA approach (Carrier sense multiple access with collision avoidance) • A potential transmitter listen to the channel for a random time in a CW to sense any ongoing transmission in progress • Channel assumed free  Transmit the packet with procedeure RTS/CTS/Data/ACK • Channel busy  Transmission postponed then node switches off the radio • Adv Flexible • Dis High latency and idle listening Random Time RTS Preamble+Data Tx Listen Sleep node1 TX CTS ACK node3 TX Rx Listen Sleep node2 Contention • The TDMA approach (Time division multiple access) • Time is divided into slots that are (in some way) assigned to neighbouring nodes • ADV: collision free and energy efficient • DIS: Low flexibility

7. Motivation for MERLIN • Due to low duty cicle of WSNs, separate MAC and Routing layers cause an extremely high latency • (e.g. SMAC and DSR  tens of seconds delay for packets of nodes in hop 10 or more) • Layer modularisation requires higher memory capability  Layer integration is beneficial • Nodes are cheap and not reliable • failure, interference, depletion, mobility  Addressing a single node can result in high error probability

8. Objectives of MERLIN • MAC+Routing integration features into a simple architecture; • No usage of handshake mechanisms; • No specific node addressing; • Reduce latency while ensuring a very low energy consumption • Controlled packet duplication to address sensor failure and bad channel condition;

9. What is the main IDEA behind the MERLIN protocol? (European EYES project, NL) Gateway Node Why Time Zones? Nodes with the same color are in the same time zone Nodes within the same subset belong to the same gateway --------------------------------- Nodes within the same zone wake up and go into sleep simultaneously

10. Division of the network in timezones SYNC packets from the gateway are forwarded to further nodes. Every node sets its zone and forward the packet to more distant nodes.

11. 4-Zone V-scheduling table Nodes in the same timezone contend the slot for local broadcast only once each 4 frametimes

12. Data traffic • Downstream multicast: Packets are transmitted to higher zones • Local broadcast: Packets reach all neighbours. No forwarding is performed • Upstream multicast: Packets are forwarded to smaller zones

13. The minimum wakeup concept through CCA Transmission Mechanism (I) • Alternation of long period of inactivity to tiny period of channel assessment; • The Clear Channel Assessment CCA is the shortest time period needed for nodes to sense any activity on the channel (~2.5msec in BMAC) • Much shorter CCA period than time required for a control packet (e.g. 35msec for 5byte transmission with Tr1001) • duty cycle reduced to less than 1% Ts Sleep period CCA Time

14. 2 questions: • How can a Tx know when the Rx is awake? • If not addressing a specific node (in multicast and broadcast), how can correct/incorrect receptions be notified?

15. Burst tone can help • Properties • Are signal impulse Do not contain any coded information • Are robusts  Several simultaneous burst can still be as one • They are shorter that a normal ACK • Utilization Multicast: Bursts identify correct receptions Broadcast: Bursts identify reception errors

16. Transmission Mechanism (II) Tx1 Tx1 Tx2 Tx2

17. Disadvantages Zone 3 Zone 5 Zone 1 Zone 2 Zone 4 A B 1)MERLIN does not address a specific receiving node  multiple copy of the same msg sent can be generated increase overhead! 2) Some collision due to the Hidden Terminal Problem (HTP) Zone 3 A ? B

18. 3 small buffers of upstram, downstream and local broadcast are provided Packets organised in multiple msgs of the same data traffic type; Packets contain a msg-ID index of included msgs; Nodes, which lose the contention, keep on listening to the beginning of the transmitted packet then go into sleep; Nodes discard from their buffer the msgs already fowarded. Routing characteristics (I) Controlled multipath Channel contention P a c k e t messages Msg-index • Pro : Reduce overhead in transmission! • Con : Small increase of node activity; • Increase complexity. Discard msgs already forwarded from their queue Listen to the packet index

19. Routing characteristics (II) Timezone maintenance • Timezone update are sent periodically; • Failed reception of timezone update from zone N-1 node to zone N node triggers a upstream multicast of Timezone Update request (TUR) • N-1 node/s reply  Connection reestablished • N-1 failed  local broadcast TUR • At least one reply  change of zoen to N+1 • N failed  downstream broadcast TUR 2 2 1 1 2 1 2 1 3 3 3 3 4 4 2 4 4 6 TUR 4 TUR 5 3

20. Phase 2: Simulation and Comparison with two existing protocol architecture: SMAC (mac)+ ESR (routing)

21. Simulation and result Nodes with the same colors are in the same zone (same hop Count Number). Number slot /frame = 4 Contention period = 30ms DataRate = 115200 bits/sec DataSize = 16+8 Bytes (data + 3 bytes preamble + starting code) Eyes node

22. V-Scheduling V scheduling Network lifetime. V-scheduling The network lifetime depends linearly on the frame length; 300 250 200 Network Lifetime (days) 150 100 50 0 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 Frametime (sec) 1 Gateway 100 Nodes rand. Distributed. 800*500 area network Min signal strength(12 m) 50 msg/min sent by 5 rand. nodes Static network • The network is considered to fail when 30% of nodes are depleted. • Lifetime calculated for a linear depletion of 2 AA batteries.

23. V scheduling setup time • V scheduling can be setup in less than 10 seconds up to 250 nodes/100m^2 of network density.

24. End-to-end packet delay V-scheduling • The controlled multiple path mechanism may cause a lower delay for nodes farther from the gateway; • An increase of latency at the intersection of data traffic flows due to periodical stop of nodes activity that go into sleep. • V-scheduling delay obtained for 2sec frametime length Frametime length should be based upon application requirements.

25. The SMAC protocol Receiver Transmitter RTS CTS time Data ACK • SMAC divides time in two periods: active time and sleeping time; • Active period = SYNC period for node sync update, Request To Send (RTS), Clear to Send (CTS). • Communication establishing: • neighboring nodes synchronize to the start of the active period then local broadcast of SYNC packets. • Data message exchanges follow the RTS/CTS/DATA/ACK; • nodes switch between different states periodically.

26. SMAC Coordinated Sleeping listen sleep listen listen t1 t2 Timing relationship of packet Tx/Rx

27. Scenario and Setup • Scenario • 5 nodes two-hops • 70 nodes Random • multihop • Metrics: • Energy consumption per RX packet • Network lifetime • E-to-E latency • Total packet overhead • % sleeping time • Parameters: • Duty cicle (acting on CW and frametime size • Low traffic conditions (12 packet/min) • High traffic conditions (60 packet/min) Forwarder Sources Destinations

28. Low traffic 2-hops scenario

29. High traffic 2-hops scenario

30. Multihop scenario: Lifetime Note: These graphs have little relevance if not related to the EtoE latency

31. Multihop scenario: Latency/energy

32. Total packet overhead The MAC routing integrated nature MERLIN results in a smaller packet overhead than SMAC+ ESR.

33. Conclusion • Description and simulated results of MERLIN have been presented; • MERLIN is suitable for large scale sensor networks with energy consumption as main goal; • MERLIN is suitable for communication to a from the gateway • The multicast mechanism with burst ACK showed large improvement on the communication reliability • The integrated nature, the absence of handshake mechanisms help reducing the EtoE packet delay • EtoE delay can be traded-off for an longer network lifetime Results showed lifetime being extended by a factor of 2.5 of MERLIN with respect to SMAC

34. Thank you for your kind attention