Maneuvering Vehicle Tracking With Bayesian Changepoint Detection

Presented is a method, referred to as the Changepoint Filter (CPF), to track dynamic maneuvering vehicles given only a set of noisy measurements. This is accomplished by utilizing a single parametrized model and detecting vehicle maneuvers with online Bayesian changepoint detection. The state of the vehicle and the changepoint probability are jointly computed at each time step. Also presented is a derivation for an exact discrete-time kinematic motion model for planar moving vehicles. This changepoint tracking method is applied with the presented planar kinematic model for both simulated and real data and the results are compared to other existing methods.