Variable Bayesian based Adaptive Cubature Kalman Filter for Vehicle Motion State Estimation under Multi Information Collaborative Perception Strategy

Abstract

Recently, the automotive autonomous driving industry has developed rapidly and there is an urgent need to update the autonomous driving technology of automobiles. The stable operation of automotive automation systems relies on the accurate perception of various information by the vehicle. Currently, information perception is affected by the complexity of road conditions and its stability and accuracy cannot meet the requirements. In view of this, the study is on the ground of a multi-information collaborative perception strategy and innovatively integrates the variational Bayesian adaptive volume Kalman filtering algorithm to estimate the vehicle motion state, with the purpose of improving the motion state estimation. Firstly, the study selected a distributed multi-information collaborative perception strategy and then, on the ground of the multi-information collaborative perception strategy, a new slope prediction model was proposed. Finally, it introduces a variational Bayesian adaptive volume Kalman filtering algorithm that can adaptively vary noise, achieving accurate estimation of vehicle motion status. The work of this article was validated and compared through experiments and the outcomes showcased that the centroid lateral deviation error under the research method was smaller, with an average error of 0.012. The research method has shown high accuracy in both straight and serpentine road conditions. The experiment validates the superiority of the research method, demonstrating its accuracy and robustness in estimating vehicle motion states and providing reliable technical support for the autonomous driving technology of vehicles.