A collaborative control framework with multi-leaders for AUVs based on unscented particle filter

In view of problems of low independent self-positioning accuracy of autonomous underwater vehicle (AUV) with low precision sensors, and the strong nonlinear characteristics of motion control model and non-Gaussian noise when filtering using the distances between leaders and followers, a collaborative localization framework with multi-leaders for AUVs is presented, which is based on the unscented particle filter. This approach lets unscented Kalman filter act on each particle of particle filter algorithm, which makes particle mix the latest posterior information of measurement when updating. It solves the depletion problem of particles to some extent and improves the effectiveness of filtering method. Collaborative location simulating experiments are carried out in two leaders model with different noises conditions. The implementation results show that this method effectively enhances the estimation accuracy of followers position.