A wavelet-based grey particle filter for self-estimating the trajectory of manoeuvring autonomous underwater vehicle

In this paper, a novel particle filter termed the wavelet-based grey particle filter (WG-PF) is proposed to self-estimate the trajectory of a manoeuvring autonomous underwater vehicle (AUV) without prior manoeuvring information. To implement the WG-PF, the particles are sampled by the state transition and grey prediction. The state transition is based on a prior dynamic model, while the grey prediction is a kind of model-free method that predicts the state through historical measurements. Therefore, the WG-PF has the inherent advantages of both model-based and model-free systems. Additionally, the measurement noise covariance is modified by the wavelet transform. Thus, the WG-PF can effectively correct the prior distribution and likelihood function of the particles and then alleviate the sample degeneracy problem. The estimation performances of three filters, the proposed WG-PF, the multiple model particle filter (PF) and the adaptive extended Kalman filter, are evaluated and compared through the experimental data, in which a trajectory plotted by the underwater acoustic positioning sensors is employed as the true trajectory. The grey-prediction-based and wavelet-based PFs are also examined to demonstrate their positive effects in the WG-PF. The presented results show that the WG-PF acquires satisfactory effectiveness, robustness and better estimation accuracy than the other filters.