Human-Like Rewards to Train a Reinforcement Learning Controller for Planar Arm Movement.
Kathleen M. JagodnikPhilip S. ThomasAntonie J. van den BogertMichael S. BranickyRobert F. KirschPublished in: IEEE Trans. Hum. Mach. Syst. (2016)
Keyphrases
- reinforcement learning
- human arm
- optimal control
- control scheme
- control policy
- adaptive control
- actor critic
- markov decision processes
- closed loop
- function approximation
- control system
- optimal policy
- reinforcement learning algorithms
- state space
- model free
- control strategies
- learning algorithm
- controller design
- reward function
- learning process
- trajectory data
- control architecture
- feedback control
- pid controller
- fuzzy controller
- parallel manipulator
- dynamic programming
- machine learning
- robotic control
- function approximators
- temporal difference
- real time
- human robot interaction
- inverted pendulum
- control algorithm
- human behavior
- reward shaping
- multi agent