Automatic driving training method, device, equipment and medium

Abstract

The invention discloses an automatic driving training method and device, equipment and a medium. The method comprises the steps that the traffic environment state at the current moment and corresponding structured noise are acquired; wherein the structured noise is determined based on historical data, the historical data is data stored in the process of pre-training the automatic driving vehicle,and the historical data comprises historical action information and historical traffic environment state information; then the corresponding execution action is determined by using the traffic environment state and the structured noise through a strategy network, and the automatic driving vehicle is controlled to execute the execution action; the strategy of the strategy network is evaluated through an evaluation network according to the execution action to obtain a corresponding return; the evaluation network parameters is updated through back propagation operation based on the return; and strategy network parameters are updated by using a strategy gradient algorithm. The stability of automatic driving training can be improved, so that the occurrence probability of dangerous accidents isreduced.

Description

technical field [0001] The present application relates to the technical field of automatic driving, and in particular to an automatic driving training method, device, equipment and medium. Background technique [0002] In modern urban traffic, the number of motor vehicles is increasing day by day, road congestion is serious, and traffic accidents occur frequently. Studies have shown that the time wasted due to traffic congestion is as long as 3 years in everyone's life, and 90% of traffic accidents are caused by human errors or mistakes. In order to minimize the harm caused by human factors, people turn their attention to the field of automatic driving. According to the degree of participation of the driver in the driving process of the vehicle, automatic driving is divided into six levels from low to high, Level-0 to Level-5, namely human driver driving, assisted driving, partial automatic driving, and conditional automatic driving , highly autonomous driving and fully au...

AI Technical Summary

Problems solved by technology

However, for DRL-based autonomous vehicles to learn autonomous driving from scratch, the selection of poor action steps in the sequence decision-making process will lead to large training variance, which is reflected in the vehicle's unsteady driving, and even accidents such as running out of the lane and colliding.

The existing research results show that compared with the modular method and the Open-loop imitation learning method, the stability of the automatic driving training based on DRL is the worst, and it is very sensitive to the environment and weather changes.

Images