Research on GFA-UFastSLAM Algorithm with Dynamic Trigger Mechanism for Autonomous Underwater Vehicles
Sun Shaohua, Zeng Qingjun, Xiaoqiang Dai, S. Le · 2025
Simultaneous localization and map construction (SLAM) is one of the important means for autonomous underwater vehicles (AUVs) to achieve navigation and localization in complex underwater environments. In this paper, a sensor fusion UFastSLAM (GFA-UFastSLAM) algorithm based on gravitational field optimization is proposed to address the problems of positional offset and cumulative error of AUV s in underwater SLAM tasks due to motion aberrations, particle degradation of sonar data, and nonlinear environments. The algorithm fuses inertial sensor (IMU) and odometer data through a multimodal innovative design to dynamically correct the motion distortion of the sonar image and eliminate the nonlinear effect of the AUV position change on the sonar data during the scanning cycle, thus dramatically improving the underwater SLAM performance. Secondly, the particle set distribution is optimized by combining the gravitational field optimization (GFA) with Dynamic Trigger mechanism in the UFastSLAM framework, which effectively suppresses the particle degradation and impoverishment problems, and the MATLAB simulation shows that the maximum error of the position estimation of the algorithm is reduced by 19.36% compared with that of the traditional particle-filtered SLAM algorithm. After the lake test experiments on the self-developed small-scale autonomous underwater vehicles, it is shown that the proposed algorithm reduces the position estimation error by 25.09% compared with the FastSLAM algorithm. The research results provide a highly robust solution for long-time autonomous navigation of AUV, which is of great value for engineering applications in resource-limited environments.