1. Energy Aware Routing for PicoRadio Rahul C. Shah Berkeley Wireless Research Center

2. Wireless Sensor Networks Dominant trend in wireless industry: More bits/sec/Hz Wireless sensor networks offer: More bits/$/nJ

3. PicoRadio System Design

4. Wireless Sensor Nodes – Constraints Low Data Rates << 10 kbps Self-configuring, maintenance-free and robust Aggressive networking protocol stack Redundancy in deployment Low cost: < 1$ Small size: < 1 cm 3 Low power/energy Long lifetime of product requires energy- scavenging Plausible scavenging level: < 100  W

5. Energy Scavenging

6. Practical Means of Energy Scavenging

7. Protocol Stack Issues at the network layer: Addressing Addressing will be class based: Symbolic addressing may be supported Routing Should route packets to the destination Given: Destination location Position of self Position of the neighbors Physical Data Link Network Application

8. Distributed Positioning [Chris Savarese(UCB)]

9. Data Link Layer Functions Transfers data between network and physical layers; Maintains neighborhood info Power control, error control and access control Computes location Controller Sensors Actuators

10. Mostly-Sleepy MAC Layer Protocols Receiving a bit is computationally more expensive than transmitting one (receiver has to discriminate and synchronize) Most MAC protocols assume that the receiver is always on and listening! Activity in sensor networks is low and randomActivity in sensor networks is low and random Careful scheduling of activity pays off big time, but … has to be performed in distributed fashionCareful scheduling of activity pays off big time, but … has to be performed in distributed fashion

11. A Reactive PicoMAC Truly Reactive Messaging Power Down the Whole Data Radio Reduce Monitoring Energy Consumption by 10 3 Times Wakeup Radio will Power Up Data Radio for Data Reception Multi-Channel Access Scheme To Reduce Collision Rate To Reduce Signaling Overhead (Shrink Address Space)

12. Multi-Channel Access Scheme SCA TCA RCA Channel Assignment Using Distributed Graph Coloring (combined with discovery) Receiver-based Channel Assignment: Channel code used as address [Chunlong Guo(UCB)]

13. Reactive Radio Issues Broadcast and data communication modes must co- exist simultaneously Sleeping nodes Communicating nodes Sleeping nodes have to wake-up to broadcast signals, and not to any signal leaking from surrounding communicating nodes Broadcast signals should not disrupt data transmission

14. PicoRadio Routing Protocol

15. PicoNetwork Specifications Density of nodes – 1 node every 1 to 20 sq. m. Radio range – 3 to 10 m Average bit rate per node ~ 100-500 bps Peak bit rate per node ~ 10 kbps Very low mobility of nodes Loose QoS requirements: Sensor data is redundant, so reliability is not required Most data is delay insensitive

16. Routing Protocol Characteristics Ensure network survivability Low energy (communication and computation) Tolerant and robust to topology changes Scalable with the number of nodes Light weight

17. Network Survivability Critical node to maintain network connectivity (network issue) Critical node as it is the only one of its type Network survivability is application-dependent – coverage may also be an issue

18. Proactive vs. Reactive Routing Proactive routing maintains routes to every other node in the network Regular routing updates impose large overhead Suitable for high traffic networks Reactive routing maintains routes to only those nodes which are needed Cost of finding routes is expensive since flooding is involved Good for low/medium traffic networks

19. Traditional Reactive Protocols Finds the best route and then always uses that! But that is NOT the best solution! Energy depletion in certain nodes Creation of hotspots in the network Source Dest

20. Directed Diffusion † Destination Source Setting up gradients Destination Source Sending data Destination initiated Multiple paths are kept alive † C. Intanagonwiwat, R. Govindan and D. Estrin, “Directed Diffusion: A scalable and robust communication paradigm for sensor networks”, IEEE/ACM Mobicom, 2000

21. Energy Aware Routing Destination initiated routing Do a directional flooding to determine various routes (based on location) Collect energy metrics along the way Every route has a probability of being chosen Probability  1/energy cost The choice of path is made locally at every node for every packet

22. Setup Phase Controller Sensor Directional flooding 10 nJ 30 nJ (0.75*10) + (0.25*30) = 15 nJ p 1 = 0.75 p 2 = 0.25 Local Rule

23. Data Communication Phase 1.0 0.6 0.4 Controller Sensor 0.3 0.7 Each node makes a local decision

24. What’s The Advantage? Spread traffic over different paths; keep paths alive without redundancy Mitigates the problem of hot-spots in the network Has built in tolerance to nodes moving out of range or dying Continuously check different paths

25. Energy Cost The metric can also include: Information about the data buffered for a neighbor Regeneration rate of energy at a node Correlation of data

26. Simulation Setup Simulations done in Opnet 76 nodes in a typical office setup 47 light sensors 18 temperature sensors 7 controllers 4 mobile nodes Light sensors send data every 10 seconds, while the temperature data is sent every 30 seconds Comparison with directed diffusion routing

27. Simulation Model Office layout Node layout Network model

28. Simulation Measurements Energy used is measured: For reception: 30 nJ/bit For transmission: 20 nJ/bit + 1 pJ/bit/m 3 Packet sizes are ~ 256 bits 1 hour simulation time Energy (mJ)Avg.Std. Dev.MaxMin Diffusion14.9912.2857.440.87 Energy Aware Routing 11.769.6741.110.98

29. Energy Usage Comparison Diffusion RoutingEnergy Aware Routing Peak energy usage was ~50 mJ for 1 hour simulation

30. Normalized Energy Comparison Diffusion RoutingEnergy Aware Routing Energy of each node is normalized with respect to the average energy

31. Bit Rate Comparison Diffusion RoutingEnergy Aware Routing Peak bit rate was 250 bits/sec. Average bit rate was 110 bits/sec.

32. Network Lifetime Nodes have fixed initial energy – 150 mJ Measure the network lifetime until the first node dies out Diffusion: 150 minutes Energy Aware Routing: 216 minutes 44% increase in network lifetime

33. Funneling Algorithm Interest Flooding Data Communication [w/ Dragan Petrović (UCB)]

34. PicoRadio Implementations

35. PicoNode I sensordigitalpowerradio Off-the-shelf fully programmable communication/computation node

36. PN3 Architecture - Rx Two Channel Channel Spacing ~ 50MHz 10kbps/channel Issues include noise suppression and isolation between RF filters Prototype Target: 3mA @ 1V RF Filter LNA f clock RF Filter Peak Det  f clock RF Filter Peak Det 

37. PN3 Architecture - Tx Use simple modulation scheme (OOK) Allows efficient non-linear PA Target output power: 0dBm Prototype Target: 4mA @ 1V PA Matching Network MOD1 MOD2 OSC1 OSC2 Preamp

38. PN3 Cycled Receiver RX0 TX0 RX1 TX1