Simultaneous numerical optimization for data association and parameter estimation

In some tracking environments, there is no clear linkage of measurements to objects, requiring data association algorithms to associate measurements to target tracks. Further complicating the problem is that targets can operate in different modes and hence have time-varying dynamics. It has recently been shown that treating the potential measurement data for an object as a single continuous signal, that contains impulses when the measurement data jumps to a new object, allows the data association problem to be cast as a trajectory optimization problem. This optimization problem can also be expanded to formulate estimates for dynamic parameters of the objects being measured. The contribution of the current paper is the development of a numerical optimization algorithm that allows the recently developed impulse-based data association and parameter estimation methods to be easily applied without a significant amount of scenario-specific analytical calculations.