Distributed target tracking in multiple widely separated radar architectures

Hana Godrich, Ali Tajer, H. Vincent Poor

Abstract: Tracking the location of a maneuvering target by a set of widely separated multiple transmit and receive radars can be done in a centralized way, where all raw measured data is communicated to a fusion center (FC), in a decentralized manner, or in a distribute approach. In this paper a distributed operational mode is proposed, in which a local tracker is implemented at each receive radar, offering redundancy, in addition to communication savings. All receivers share the set of available transmit antennas. The information exchanged among the local trackers, or local FCs, is limited to the predicted target state. Each local predictor is updated based on predictions collected from all other FCs, treated as a set of experts' advice. An exponential weighted average forecaster is derived, resulting in a local prediction loss that is upper-bounded by the best expert performance with a minimal cost in term of information exchange.