PIAPF for manoeuvring target tracking in the presence of multiplicative noise

Abstract

In this study, a prediction-interval-based adaptive particle filter (PIAPF) is developed to track a manoeuvring target in the presence of multiplicative measurement noise. In PIAPF, input augmentation technique is utilised to estimate the state variables of target and unknown inputs (manoeuvres) simultaneously. To cope with unknown sudden changes of system state variables (caused by manoeuvres), the covariance matrix of the importance density function is adaptively adjusted based on the prediction interval of the output estimation. In addition, a theorem is developed which confirms that the output estimation error is upper bounded by a given probability. The likelihood function of the non-stationary state-dependent error of sensors, which is modelled as multiplicative noise, is then obtained for weight calculation of PIAPF. The proposed PIAPF is then used to track a manoeuvring target in a wireless sensor network with distance-measuring sensor nodes. Simulation results demonstrate the effectiveness of the proposed PIAPF in terms of tracking accuracy and computational load.