Abstract: In order to solve the problem of maneuvering object tracking by autonomous mobile robots in unknown environments, a filtering algorithm adopting probabilistic data association (PDA) and interacting multiple model (IMM) is proposed. The algorithm is based on the framework of full covariance extended Kalman filter. The robot state, environment landmark states and target state are used to form the system state, and the problem of model uncertainty in the process of object moving is solved by IMM filter. For the problem of false object observations in practical application, a PDA method is used to weight the contribution of different observations to system state update in different moving model filters. Simulation results show the accuracy of the algorithm in estimating robot state, environment landmark states and target state, and prove the algorithm ability of tracking maneuvering object, as well as the ability of dealing with false object observations. A real robot experiment verifies the practicability of the algorithm.