Autonomous lane keeping based on approximate Q-learning

Obstacle avoidance is one of the most important problems in autonomous robots. This paper suggests a collision avoidance system using reinforcement learning. Hand-crafted features are used to approximate Q value. With off-line learning, we develop a general collision avoidance system and use this system to unknown environment. Simulation results show that our mobile robot agent using reinforcement learning can safely explore a corridor even if the agent does not know the shape of corridor at all.

[1]  Min Guo,et al.  Reinforcement Learning Neural Network to the Problem of Autonomous Mobile Robot Obstacle Avoidance , 2005, 2005 International Conference on Machine Learning and Cybernetics.

[2]  Xin Xu,et al.  Dynamic path planning of a mobile robot with improved Q-learning algorithm , 2015, 2015 IEEE International Conference on Information and Automation.

[3]  Francesco Borrelli,et al.  An auto-generated nonlinear MPC algorithm for real-time obstacle avoidance of ground vehicles , 2013, 2013 European Control Conference (ECC).

[4]  Peter Dayan,et al.  Q-learning , 1992, Machine Learning.

[5]  Tai Lei.  A Robot Exploration Strategy Based on Q-learning Network , 2016 .

[6]  Richard S. Sutton,et al.  Reinforcement Learning: An Introduction , 1998, IEEE Trans. Neural Networks.

[7]  Xiaogang Ruan,et al.  Application of reinforcement learning based on neural network to dynamic obstacle avoidance , 2008, 2008 International Conference on Information and Automation.

[8]  He Bing,et al.  A route planning method based on improved artificial potential field algorithm , 2011, 2011 IEEE 3rd International Conference on Communication Software and Networks.

[9]  V. Madhu Babu,et al.  An autonomous path finding robot using Q-learning , 2016, 2016 10th International Conference on Intelligent Systems and Control (ISCO).

[10]  Shane Legg,et al.  Human-level control through deep reinforcement learning , 2015, Nature.