Optimal Sensor Trajectories in Bearings-Only Tracking

We consider the problem of determining sensor trajectories in the bearings-only tracking of an uncertain target. This work differs from previous research in that we allow the target dynamics to be both uncertain and random; and also consider multi-sensor scenarios. The basis of our technique is to control a measure of estimation error based on the Posterior Cramér-Rao lower bound (PCRLB), and we present both single-step and multi-step planning approaches. We also introduce an efficient search technique that allows us to quickly perform the necessary optimisation(s). Sensor trajectories are shown to be almost identical to those obtained by performing an enumerative search, but the computational load is reduced by several orders of magnitude. Simulation results are presented for 1, 2 and 3 sensor scenarios, and compare the single-step and twostep approaches. The performance (in terms of both the PCRLB and particle filter estimation errors) improves both with two-step planning and as sensor numbers increase, as one would expect.