Research On Virtual Link Simulation And Driving Behavior Robot Based On DDQN Algorithm

With the development of computer and big data,artificial intelligence and reinforcement learning have also been developed.Automatic driving technology has also been a certain attention and development.This requires a simulation environment that can provide virtual links for the reinforcement learning algorithm,which is very important for training the virtual driving behavior robot,creating a good simulation environment and training it with reinforcement learning algorithm,therefore,creating a good virtual link simulation environment has been more and more people’s attention,based on this,this paper created a virtual link simulation training environment,the reinforcement learning algorithm based on DDQN is used to test the rationality of environment design and congestion algorithm.Reinforcement learning algorithms are part of a large family of machine learning algorithms.Unlike supervised or unsupervised learning,which has a large amount of experience or input data,they are basically self-taught.Deep learning has strong perception ability,but lacks certain decision-making ability.Reinforcement learning,however,has the ability to make decisions and can do nothing about perception problems.Therefore,the combination of the two can be regarded as deep reinforcement learning,complementing each other’s advantages,and providing a solution to the perceptual decision-making problem of complex systems.In real life,when we drive and travel in daily life,we will be affected by various factors.Under the influence of various factors,human beings will become more and more proficient in driving behavior and have higher and higher driving level through continuous learning and memory.Similarly,the driving behavior robot is in the same situation.By constantly rewarding and punishing the robot based on its driving behavior,it can learn how to drive more accurately.The main research contents of this paper are as follows:(1)Based on the analysis of people’s driving situations in daily life,a comprehensive robot simulation environment of virtual link driving behavior is constructed and corresponding rewards are set.(2)The design of congestion algorithm and the detailed design of congestion algorithm.(3)Training driving behavior robot based on DDQN reinforcement learning algorithm.