Research On Collaborative Tracking Method For Intelligent Connected Vehicle Based On Bayesian Filter

Intelligent Connected Vehicle(ICV)relies on perception,decision-making,and control technology to achieve autonomous driving.Perception technology,as the first step of autonomous driving,is the foundation of autonomous driving decision-making and control.ICV with highprecision perception ability has a higher level of safety.Target tracking is a key technology in environmental perception.Due to factors such as obstacles and sensor installation angle limitations,traditional single vehicle autonomous tracking has problems such as low tracking accuracy and large blind spots.With the continuous development of communication technology,ICVs can exchange perception data through workshop communication to achieve collaborative tracking.This thesis systematically studies collaborative tracking methods for target vehicle motion uncertainty,inaccurate collaborative vehicle positioning,and optimal collaborative vehicle screening.This provides theoretical basis and technical support for ICV to achieve more accurate decision control.The main work is as follows:(1)A collaborative tracking method is proposed to address the issue of dynamic changes in the motion model of the target vehicle caused by its maneuverability.Combining improved interactive multiple models and particle filter algorithms,an adaptive interactive multiple model particle filter algorithm that can adapt to the dynamic changes of maneuvering target motion models was derived.A collaborative tracking framework was designed for target layer fusion,and data association and fusion were completed using the Hungarian algorithm and the fast covariance crossover algorithm.The simulation results show that the proposed collaborative tracking method can significantly improve the tracking accuracy of maneuvering targets without affecting tracking efficiency.(2)A self-adaptive collaborative tracking method is proposed to jointly estimate the target vehicle state and the probability of collaborative vehicle positioning loss in response to the inaccurate positioning information of collaborative vehicles.Derived an adaptive Kalman filtering algorithm based on variational Bayes,and designed a collaborative tracking framework for data layer fusion.The host vehicle fused the perception data of the collaborative vehicle,and completed the collaborative tracking by executing adaptive Kalman filtering based on variational Bayes.The simulation results show that the proposed collaborative tracking method can accurately estimate the localization loss probability of the collaborative vehicle,significantly improving the tracking accuracy of the system,and has better adaptability when the localization loss probability of the collaborative vehicle is unknown and time-varying.(3)A collaborative tracking method based on the optimal collaborative vehicle screening strategy is proposed to address the issue of how to select the optimal collaborative vehicle.The impact of four factors,ICV distance,ICV quantity,GPS accuracy,and Lidar accuracy,on collaborative tracking was analyzed and verified.Based on the analysis results,the weights of different influencing factors were calculated using the hierarchical entropy weight subjective and objective comprehensive weighting method,and the expression of trust was obtained.Collaborative vehicles were selected for collaborative tracking according to the trust level of different ICVs in the environment.The simulation results show that the proposed collaborative tracking method can balance tracking accuracy and efficiency,significantly improving tracking accuracy while reducing and meeting the requirements of online real-time tracking.