Distributed data association for multiple-target tracking using game theory

In this paper, we develop a game-theoretic framework to address data association for multiple-target tracking problems. We model the interaction among trackers as a game, by considering them as players, and the set of measurements as strategies. We develop utility functions for the players, and use a regret-based learning algorithm to find the equilibrium of the game. We will then use Monte Carlo filters, operating in parallel, to track state vectors corresponding to the individual targets. In contrast to the traditional Monte Carlo filters that sample the association vector, we first find the association in a deterministic fashion, and then use Monte Carlo sampling on the reduced dimensional state of each target independently, thereby enabling a distributed implementation. We provide numerical results to demonstrate the performance of our proposed filtering algorithm.