Improving state-action space exploration in reinforcement learning using geometric properties.
Ion MateiRaj MinhasJohan de KleerAnurag GanguliPublished in: CDC (2017)
Keyphrases
- geometric properties
- state action
- space exploration
- reinforcement learning
- evaluation function
- markov decision process
- action space
- state space
- function approximation
- function approximators
- obstacle avoidance
- average reward
- state transitions
- stochastic games
- optimal policy
- reinforcement learning algorithms
- geometric constraints
- machine learning
- kernel matrix
- optimal control
- belief state
- model free
- action selection
- markov decision processes
- real valued
- fuzzy logic
- computer vision
- learning algorithm