Regularizing Action Policies for Smooth Control with Reinforcement Learning.
Siddharth MysoreBassel MabsoutRenato MancusoKate SaenkoPublished in: ICRA (2021)
Keyphrases
- reinforcement learning
- control policy
- control policies
- optimal policy
- action selection
- control problems
- action space
- optimal control
- fitted q iteration
- robot control
- control method
- learning algorithm
- partially observable domains
- reinforcement learning algorithms
- model free
- markov decision process
- control system
- reward shaping
- multi agent
- neural network
- reward function
- temporal difference
- function approximation
- partially observable markov decision processes
- control strategy
- transfer learning