1. Mohamed Hauter CMPE 259 – Sensor Networks UCSC 1

2. * Introduction * Objectives * Proposals and approaches * Related Work * Simulations and Results * Strengths and weaknesses 2

3. * An energy-efficient sensor network * Minimal number of sensor nodes in active mode * Increase the lifetime of the sensor network * Prevent connection degradation 3

4. * Terminology: * DPM: Dynamic Power Management * OGDC: Optimal Geographical Density Control * ACPI: Advanced Configuration and Power Interface 4

5. * Approach: * Tackle energy efficiency on all levels of the entire network * Dynamic power management = shutting down nodes when not needed and wake them up when necessary * Consideration of the state of components ( microprocessor, A/D converter, memory, transceiver, etc.) when making a decision to turn off a node 5

6. * Approach (continue): * Density control while maintaining: a. Coverage b. Connectivity * Localized density control algorithm 6

7. * Approach (continue): * Consideration of battery status and energy wasted in the process of node-awakening * Incorporate OGDC in the control logic 7

8. * Verity of DPM techniques * Dynamic Voltage Scaling * Dynamic Voltage and Frequency Scaling * Sentry based power management (application driven) * Software and operating system power management 8

9. * Weaknesses of traditional predictive techniques: * Cannot provide an accurate tradeoff between energy saving and performance degradation * Does not deal with systems in which requests can be queued 9

10. * Power aware sensor node model: * Node components: processor, memory, AD converter, and transceiver (radio) * Components of each node can be in different states: active, idle, or sleep * Different combinations of component power modes 10

11. 11

12. * Sleep-state transition policy: * P = Power Consumption * t = Time of event * s = sleep state * Tau = transition mode 12

13. * System Parameters: 13

14. * 50x50 meters area of coverage * 100 nodes * Uniformly and randomly distributed * Nodes are capable of directly communicating with the host * Each nodes initial energy is 100 joules 14

15. 15

16. Strengths: * An energy-efficient sensor network * Minimal number of sensor nodes in active mode * Increase the lifetime of the sensor network * Prevent connection degradation Weaknesses: *Analysis did not take latency into account * Events missed during deepest-sleep state * OGDC requires knowledge of nodes location (extra processing and memory overhead) 16

17. * Utilize natural sources of energy (solar, motion, vibration, etc.) to recharge nodes batteries * Employ energy-saving mechanisms * Determine the sleep and wake up probabilities of nodes using a bargaining game 17

18. 18

19. 1.Solar 2. Thermoelectric 3. Vibration Based 19

20. Two types of buffers: 1.Local buffer: gathers data collected locally (through sensors). 2. External buffer: gathers data from other nodes to be relayed. 20

21. To find the optimal path to deliver data packets while considering: 1. Energy level 2. Path length 3. Path reliability Avoid: 1. Idle listening 2. Overhearing 3. Packet collisions 21

22. Using Explicit Signaling: A node notifying the access point that it is going into power-saving (PS) mode Dual Channel MAC Protocols (Avoid Collisions): Signaling channel Data transmission channel 22

23. Lazy packet-scheduling scheme Determine beginning and duration of transmission Transmit at a low data rate Save energy Packet delay and reduced throughput Tradeoffs! 23

24. QoS vs. Energy Constrains Energy harvesting limitations Integration of energy harvesting techniques across layers 24

25. Radio modes: Active – 25mW Listen – 14mW Sleep – 0.01mW Channel and queue-aware strategy Radio - Listen when queue is empty Sensor – sleep when channel quality is bad 25

26. Players: Player 1: node Player 2: data receiving entity Strategy: Player 1: select wakeup probability when in sleep mode Player 2: select wakeup probability when in listen mode Payoff: Player 1: packet blocking probability Player 2: packet dropping probability 26

27. 27

28. 28

29. 29

30. Strengths: 1. Energy efficient 2. Incorporates the states of different components of the network Weaknesses: 1. battery energy level is not taken into consideration when making a sleep/wakeup decisions 2. Data transmission delay – low data transmission rates 3. The assumption of one-hop routing model in which all nodes can reach the sink is not practical 30

31. * Prolong the lifetime of the network * Minimizing the data processing and communication costs * Employ multi-hop communications effectively 31

32. * LEACH: dividing the sensor network into cluster heads (CH) which can communicate with sinks and amongst themselves. Cluster Heads are constantly changing (random selection) to prevent draining its energy. * SOP: a tree of cluster heads is built using fixed nodes. * EDETA: builds a hierarchal tree among cluster heads to avoid direct communication with sink. 32

33. * HARP can save more energy by forming intra- cluster hierarchal architectures in conjunction with inter-cluster trees. * Leverage node mobility to enhance network performance in terms of coverage, lifetime, energy efficiency, and latency. 33

34. * Two hierarchal tree structure: 1. Between CHs and the sinks 2. Within the cluster * HARP has a local reconfiguration scheme in case of a failure * Supports more than one sink - scalability 34

35. 35

36. 36

37. 37

38. 38

39. * Causes of failure: * Battery depletion, node malfunction, multipath fading, low link quality, or node mobility. * Mechanisms; 1. The recovery slot 2. The substitute node 39

40. * Unlike the LEACH approach, HARP ensures that nodes all die at the same time * Solves the problem of the extra energy waste of CHs * CHs are randomly selected, unless new node has less energy than existing CH. 40

41. 41

42. 42

43. 43

44. * Strengths: * Very high level of energy efficiency * Scalable design * Efficient Local recovery capability * Optimizes routing of both upstream and downstream traffic flows * Weakness: * Increased complexity in terms of resource scheduling and network topology management * Increased memory overhead 44

45. 45