Minimized Euclidean error data association for multi-target and multisensor uncertain dynamic systems

In this paper, data association and tracking problems with multi-target and multisensor uncertain dynamic systems are considered. The methods developed in [1] for state estimation of single target systems will be extended to data association and tracking for multi-target systems in terms of minimizing Euclidean/absolute error. Assume that a nominal system model, bounds of parameters uncertainty/biases and noises are known. This type of uncertain models have also many applications. For example, uncertain biases of measurements and time stamps may be described by a bounded set. Obviously, this uncertainty framework is significantly different from that of the combination of IMM and JPDA estimators. The latter assumes that a true target model is one of several possible precise maneuvering models given the transition probabilities among these models and probability density functions of all model noises. However, the former only knows that the true model is an element of a bounded uncertain model set so that there are infinite model candidates. Besides, the optimization criterion for the latter is conventional MSE of the state estimation, but the former is to minimize Euclidian error. Clearly, removing model uncertainty or biases requires enough well-associated measurement data in advance. However, to obtain a good data association, one has to well estimate and remove the model uncertainty or biases. Since the two problems are mutually dependent and influenced, such data association and estimation problems cannot be solved well by the existing data association methods. In this paper, two minimized Euclidean-error data association (MEEDA) algorithms for single sensor and multi-sensor systems are proposed respectively. Quite a few numerical examples are given to reveal the major factors influencing the performance of MEEDA algorithms.