Autonomous navigation based on improved unscented FastSLAM algorithm

Simultaneous localization and mapping (SLAM) is an advanced method to solve autonomous robot navigation in unknown environment. In order to reduce the accuracy of AUV Unscented FastSLAM estimation caused by resampling, this paper developed an improved Unscented FastSLAM algorithm based on adaptive fading unscented particle filter. The algorithm merged the unscented particle filter with the fading filter to form adaptive proposed distribution function, and only resampled some particles whose weights are unstable. By these two aspects could make the system highly adaptive, while mitigating the degradation of particles. Simulation experiments show that compared with Unscented FastSLAM algorithm, the proposed algorithm can achieve higher SLAM estimation accuracy with fewer particles, which greatly reduces the complexity of SLAM algorithm.