Integrated AISGP Model for Real-Time Autonomous Vehicle State Estimation and Path Tracking Control

Autonomous vehicles often face issues such as inaccurate state estimation and insufficient path tracking precision caused by model biases and sensor noise in complex dynamic scenarios. Therefore, this paper proposes an autonomous vehicle state estimation and path tracking control model based on the adaptive incremental sparse spectrum gaussian process (AISGP), which integrates model predictive contour control (MPCC) and moving horizon estimation (MHE). By constructing a hierarchical AISGP model and combining a residual prediction model, the model can efficiently capture unmodeled vehicle dynamic characteristics and noise interference. The AISGP-MHE module optimizes the state estimation of lateral velocity and yaw rate, and the AISGP-MPCC module optimizes the path tracking control strategy by minimizing the contour and lag error. Evaluations on CarSim demonstrate that the proposed method improves vehicle tracking accuracy by 17% compared with traditional schemes, and the inference speed of AISGP is 8.4 times faster than that of Gaussian Process (GP).