Distributed Target tracking and localization in multi-hop networks

We consider the problem of distributed target tracking under communication constraints between the sensor nodes, a problem that has recently received significant attention. On account of communication constraints the problem necessitates not only the dynamic selection of optimal sensor nodes but also corresponding fusion centers to enable local processing of sensor information. This coupled problem is generally intractable and significant effort has been devoted towards proposing greedy strategies under various performance criteria. In contrast our paper decouples the three aspects, namely non-linear partially observed state estimation, sensor selection and fusion center locationing, and adopts a certainty equivalence perspective. The main advantage of this approach is that since significant communication costs in target tracking with multi-hop networks arises primarily from the switching of fusion centers, this problem can be isolated and optimized. In particular, we show that optimal tracking algorithms that jointly optimizes the average multi-hop communications as well as average tracking error can be derived in an infinite horizon setting. Specifically, when target dynamics is described by a random walk, the optimal strategy exhibits a hybrid switching strategy, whereby active sensor locations are held stationary until the target moves outside a threshold radius around the sensors. This holds even when there are sensors located close to the target. Surprisingly this result is fundamental to the multi-hop communication penalty and does not hold for penalties that reflect free space attenuation.