摘要
The path planning for autonomous vehicles is a hot topic in academic world.The goal of this problem is to design a vehicle with learning ability to approach the target without any collision.In this paper,we focus on data-driving reinforcement learning(RL) design for the path planning for autonomous vehicles problems.We proposed the method of sensor detection,and a self-learning strategy for a vehicle seeking the target with obstacle avoidance.Specifically,we designed a continuous reinforcement signal to improve the system's preferential decision between the target seeking and the obstacle avoidance.To verify the learning ability of our strategy,we developed an in- door environment with different experiments.The simulation results show that the RL presents an effective learning ability for the path planning for autonomous vehicles problems.
The path planning for autonomous vehicles is a hot topic in academic world.The goal of this problem is to design a vehicle with learning ability to approach the target without any collision.In this paper,we focus on data-driving reinforcement learning(RL) design for the path planning for autonomous vehicles problems.We proposed the method of sensor detection,and a self-learning strategy for a vehicle seeking the target with obstacle avoidance.Specifically,we designed a continuous reinforcement signal to improve the system's preferential decision between the target seeking and the obstacle avoidance.To verify the learning ability of our strategy,we developed an in- door environment with different experiments.The simulation results show that the RL presents an effective learning ability for the path planning for autonomous vehicles problems.
引文
[1]S X Yang,M Meng.An Efficient Neural Network Approach to Dynamic Robot Motion Planning[J].Neural Networks,2000,13:143-148.
[2]A Zhu,S X Yang.Neurofuzzy-Based Approach to Mobile Robot Navigation in Unknown Environments[J].IEEE Transaction on Systems,Man and Cybernetics,Part C;Application and reviews,2007,37(4):610-621.
[3]X Fang,X Chen,H Gao,et al.Self-driving Car Navigation with Obstacle Avoidance Using Reinforcement Learning[C].SAE-China Congress Proceedings,2015:498-501.
[4]R S Sutton,A G Barto.Reinforcement Learning:An Introduction[M].Cambridge,MA;MIT Press,1998.
[5]LP Kaelbling,ML Littman,A W Moore.Reinforcement Learning:A Survey[J].Journal of Artificial Intelligence Research,19%,4:237-285.
[6]X Fang,D Zheng,H He,et al.Data-driven Heuristic Dynamic Programming with Virtual Reality[J].Neurocomputing,2015,166;244-255.
[7]Z Ni,X Fang,H He,et al.Real-time Tracking on Adaptive Critic Design with Uniformly Ultimately Bounded Condition[C].IEEE Symposium on Adaptive Dynamic Programming and Reinforcement Learning(ADPRL'13),IEEE Symposium Series on Computational Intelligence(SSCI),2013:39-46.
[8]X Fang,D Zheng.Goal Seeking of Autonomous Mobile Robot with Obstacle Avoidance using Heuristic Dynamic Programming[J].Transactions of the Chinese Society for Agricultural Machinery,2014,45(7):73-78.
[9]X Fang,H He,Z Ni,et al.Learning and Control in Virtual Reality for Machine Intelligence[C].IEEE International Conference on Intelligent Control and Information Processing,2012:63-64.
[10]J Si,Y Wang.On-line Learning Control by Association and Reinforcement[J].IEEE Transactions on Neural Networks,2001,12(2):264-271.
[11]X Fang,D Zheng.Control Strategy Design for Car Autonomous Navigation using Adaptive Dynamic Programming[J].Journal of Yanshan University,2014,38(1):57-65.
[12]H He,Z Ni,J Fu.A Three-network Architecture for On-line learning and Optimization Based on Adaptive Dynamic Programming[J].Neurocomputing,2012,78(1):3-13.
[13]J Si,AG Barto,W B Powell,et al.Handbook of Learning and Approximate Dynamic Programming[M].IEEE Press,2004.