1 Energy-Quality Tradeoffs for Target Tracking in Wireless Sensor Networks Sundeep Pattem, Sameera Poduri, and Bhaskar Krishnamachari 2nd Workshop on Information Processing in Sensor Networks, IPSN, April 2003.

2 Outlines Introduction Quality Metric : Tracking Error Energy Metric : Tracking Energy Tracking Strategies Experiments Conclusions

3 Introduction The various possible activation strategies that illustrate the possibility of energy savings at the cost of reduced tracking quality (1) naive activation (2) randomized activation (3) selective activation based on prediction (4) duty-cycled activation

4 Introduction(Cont1.) The impact of the energy-quality tradeoffs deployed/activated density of sensors the sensor ’ s sensing range capabilities of activated and un-activated nodes the target ’ s mobility model

5 Introduction(Cont2.) a sensor network consisting N nodes operating for a total time duration T only a single target moving fixed sensing range S 2 different operation modes for sensor tracking/sensing mode use a higher power H Can sense a target and communicate with neighbor nodes communication mode use a lower power L can only communicate with neighbor nodes

6 there are k sensors at locations detecting the target at time t.

7 Quality Metric : Tracking Error the average total energy expenditure P (averaged over a period of time T ) is the actual position of the target at time t the instantaneous tracking error metric

8 the time T average error : assume that the target ’ s movement is an Ergodic random process the tracking error metric the expected distance between the estimated and actual positions of the target:

9 Energy Metric:Tracking Energy the number of nodes that are in tracking/sensing mode : ns the number of nodes that are in communication mode : nc = N − ns The average energy expenditure for a network of N nodes

10 the sensing power expenditure as being a power law function of the sensing range S of the nodes: :the decay exponent for the for the sensed signal the energy metric is

11 Tracking Strategies Naive activation (NA) all nodes in the network are in tracking mode all the time Randomized activation (RA) each node is on and in tracking mode with a probability p

12 Selective activation based on prediction (SA) only a small subset of all the nodes are in tracking mode at any given point of time Xa : the actual position of the target Xb = Xs the belief position of target as before the new belief location Xb(t + 1) Xp : the predicted target position predict the next location of the target is at Xp(t + 1)

13 ρ is the density of deployment

14 Sp S

15 Duty-cycled Activation (DA) the entire sensor network periodically turns off and on with a regular duty cycle it can actually be used in conjunction with any other activation strategy (including NA, RA and SA) the period of the cycle the on-time the average number of tracking sensors

16 Experiments 200 unit x 200 unit area with random placement of sensors density of deployment ρ= 1 sensor/unit area (i.e. a total of nodes) the target motion is linear, sinusoidal and other reasonable trajectories

17

18

19

20

21 selective activation with reasonably high Sp is a dominating, Pareto-optimal strategy

22 Performance of Duty-Cycled Activation v : the mean target speed : the instantaneous tracking error during time for the network without duty-cycling the tracking error at time t is the average tracking error for duty-cycled activation

23

24 For the same ratio tON/TD, the average tracking error Q increases with the period TD

25 The duty-cycled activation is a flexible and efficient mechanism for tuning the energy-quality tradeoff of tracking.

26 Conclusions Identify 4 generic sensor activation strategies for target tracking. Develope simple metrics to evaluate the performance of 4 strategies.