Non-linear estimation in bearings only tracking of manoeuvring targets

The problem investigated concerns the estimation of the parameters which describe the motion of a target vehicle when given bearing measurements from a moving observer vehicle. The target vehicle is assumed to be capable of abrupt changes in its heading, of an unknown magnitude, and occurring at unknown times. The target states are not directly observable from a sequence of noisy measurements and an observer manoeuvre is required in order to achieve a unique solution. This combination of a delayed observability, non-linear parameter estimation and an abruptly changing system, makes the bearings-only tracking problem particularly troublesome. In this type of non-linear filtering problem a suitable formulation can be difficult to design. The author has developed an adaptive Gaussian sum technique which `manages´ the filter bank and prevents the computational burden from becoming excessive whilst achieving the desired tracking ability. It should be noted that because of the range/manoeuvre ambiguity which occurs in this application, methods which employ `moving banks´ are inappropriate since they assume a unimodal pdf for τ