TY - GEN
T1 - Learning Control for Air Hockey Striking Using Deep Reinforcement Learning
AU - Taitler, Ayal
AU - Shimkin, Nahum
N1 - Publisher Copyright:
© 2017 IEEE.
PY - 2017/7/1
Y1 - 2017/7/1
N2 - We consider the task of learning control policies for a robotic mechanism striking a puck in an air hockey game. The control signal is a direct command to the robot's motors. We employ a model free deep reinforcement learning framework to learn the motoric skills of striking the puck accurately in order to score. We propose certain improvements to the standard learning scheme which make the deep Q-learning algorithm feasible when it might otherwise fail. Our improvements include integrating prior knowledge into the learning scheme, and accounting for the changing distribution of samples in the experience replay buffer. Finally we present our simulation results for aimed striking which demonstrate the successful learning of this task, and the improvement in algorithm stability due to the proposed modifications.
AB - We consider the task of learning control policies for a robotic mechanism striking a puck in an air hockey game. The control signal is a direct command to the robot's motors. We employ a model free deep reinforcement learning framework to learn the motoric skills of striking the puck accurately in order to score. We propose certain improvements to the standard learning scheme which make the deep Q-learning algorithm feasible when it might otherwise fail. Our improvements include integrating prior knowledge into the learning scheme, and accounting for the changing distribution of samples in the experience replay buffer. Finally we present our simulation results for aimed striking which demonstrate the successful learning of this task, and the improvement in algorithm stability due to the proposed modifications.
UR - http://www.scopus.com/inward/record.url?scp=85046676745&partnerID=8YFLogxK
U2 - 10.1109/ICCAIRO.2017.14
DO - 10.1109/ICCAIRO.2017.14
M3 - Conference contribution
AN - SCOPUS:85046676745
T3 - Proceedings - 2017 International Conference on Control, Artificial Intelligence, Robotics and Optimization, ICCAIRO 2017
SP - 22
EP - 27
BT - Proceedings - 2017 International Conference on Control, Artificial Intelligence, Robotics and Optimization, ICCAIRO 2017
PB - Institute of Electrical and Electronics Engineers
T2 - 2017 International Conference on Control, Artificial Intelligence, Robotics and Optimization, ICCAIRO 2017
Y2 - 20 May 2017 through 22 May 2017
ER -