Design And Implementation Of Self-driving Decision-making Method Based On Vehicle-road Collaboration

With the continuous improvement of social and economic development level,the number of cars in China is increasing year by year,and the phenomenon of traffic congestion in cities has become the norm.On the other hand,more than one million people die in man-made traffic accidents every year,so driving safety cannot be ignored.Automatic driving technology is one of the fundamental solutions to the above problems.At present,researchers use deep learning technology to train networks to learn human driving experience to gain the ability to drive autonomously.This research approach has achieved some results,but still faces problems such as decision delay and decision failure caused by incomplete cognition of environmental information.With the rapid development of Internet of vehicles technology,autonomous driving based on vehicle-road collaboration has become one of the development directions of autonomous driving technology.Based on this background,this paper studies the following aspects:1.This thesis proposes a perception-decision network model that supports multi-vehicle collaboration.This model can reduce the problem of decision-making failure caused by incomplete bicycle cognition by integrating the results of multi-vehicle collaborative cognition to make driving decisions.The feasibility of the model is verified by experiments.Compared with the existing end-to-end method,the model proposed in this paper has higher accuracy of corner prediction.2.Set up based on robot system of smart car simulation experiment platform,through the establishment of local area network environment simulation road collaborative environment,using a laboratory fixed server as the edge of the compute nodes,the algorithm for automatic driving training,decision on the edge,to control execution in vehicle automated driving system,completed based on collaborative end-to-end automated driving vehicle road simulation experiments.Experiments verify the feasibility and effectiveness of the end-to-end autonomous driving scheme based on vehicle-road collaborative environment.