Abstract
Accurate sensor signals critically determine the stability of intelligent vehicles. However, variations in CAN network load ratio (CANLR) inevitably introduce time-varying sensor delay (TVSD) in signal transmission, compromising state estimation reliability. To address this challenge, this paper proposes a novel RBF time delay adaptive square root Cubature Kalman Filter (RBF-TD-ASRCKF). Firstly, a three-degree-of-freedom nonlinear vehicle model is established to accurately characterize vehicle dynamics. Additionally, a distribution model relating CANLR to TVSD is derived. Our novel approach utilizes a state-augmentation method to integrate the delay into the system state, combined with a square root Cubature Kalman Filter framework to enhance nonlinear prediction accuracy. To further suppress noise interference, an RBF neural network adaptive law is designed to adjust the system noise covariance matrix in real-time. Finally, leveraging the derived delay distribution characteristics, an RBF-TD-ASRCKF is designed. This integrated methodology significantly enhances the algorithm's responsiveness to dynamic CANLR fluctuations. Hardware-in-the-Loop (HIL) testing demonstrates that the proposed scheme effectively improves state estimation accuracy for intelligent vehicles operating in complex communication environments.
Funding Information
  • National Natural Science Foundation of China (U24A20199)
  • National Natural Science Foundation of China (52372390)
  • Youth Foundation of Jiangsu Province (BK20241419)

This publication has 31 references indexed in Scilit: