This paper discusses asynchronous distributed inference in object tracking. Unlike many studies, which assume that the delay in communication between partial estimators and the central station is negligible, our study focuses on the problem of asynchronous distributed inference in the presence of delays. We introduce an efficient data fusion method for combining the distributed estimates, where delay in communications is not negligible. To overcome the delay, predictions are made for the state of the system based on the most current available information from partial estimators. Simulation results show the efficacy of the methods proposed.
Keywords: Kalman filter; data fusion; delay; distributed learning; object tracking; partial estimators; random access; statistical inference; wireless sensor network.