Practical Approaches To Distributed Tracking In Sensor Networks

Modern sensor networks are composed of both stationary and dynamic nodes which depend heavily on the sharing of information in order to optimize performance, manage resources, and maximize overall capability and capacity. When information is shared within any sensor network, the overall performance hinges on its ability to utilize and integrate the information received into the distributed solution at each node. For a sensor network, one of the primary canonical problems is target tracking, which involves solving the detection, association, and estimation problems in a distributed manner using scalable algorithms which converge in time to the optimal or near-optimal solution. In a generic sensor network, each node performs local sensing and tracking and also receives measurement and uncertainty data from other nodes. One approach for obtaining practical solutions to these types of problems is to employ distributed extended Kalman information filters (EKIF) at each node in order to perform both local and global tracking and sensor fusion. Another approach which has received extensive coverage in the current literature employs a distributed consensus protocol for obtaining the global tracking and sensor fusion solution. The true elegance of a distributed consensus approach is that all nodes composing a sensor network can be guaranteed to converge asymptotically to the identical solution. The primary target tracking performance objective for the sensor networks examined herein is achieving accurate, communal situational awareness at each node. Simulation results will be presented which demonstrate the use of distributed, extended Kalman information filters and asynchronous consensus filters in a practical sensor network.