Maneuvering target tracking using q-learning based Kalman filter

In this paper, we propose a new algorithm based on the combination of the well-known Kalman Filter (KF) and a Temporal Differencing (TD) method named the Q-learning technique. The purpose is to solve a single maneuvering target tracking problem. The algorithm consists of adapting the noise process of the Kalman filter when the target changes its mode using Q-learning based decision. After a comparison with an ideal Kalman filter and the IMM filter, the approach has shown interesting behavior. Monte-Carlo simulation results are provided to assess these outcomes.