Distributed Bayesian Filters for Multi-Vehicle Network
Our work presents a measurement dissemination-based distributed Bayesian filtering (DBF) approach for a network of unmanned ground vehicles (UGVs). The DBF utilizes the Latest-In-and-Full-Out (LIFO) local exchange protocol of sensor measurements for data communication within the network. Each UGV under LIFO only exchanges with neighboring UGVs a full communication buffer consisting of latest available measurements. Under the condition of fixed and undirected topology, LIFO can guarantee non-intermittent dissemination of all measurements over the network within finite time.
Two types of LIFO-based DBF algorithms are developed and utilized to estimate individual probability density functions (PDF) for a static target and for a moving targets, respectively. For the static target, each UGV locally fuses the newly received measurements while for the moving target, a set of measurement history is stored and sequentially fused. The consistency of LIFO-based DBF is proven by the fact that the estimated target position converges in probability to the true target position.