Hysteresis-Based RL: Robustifying Reinforcement Learning-based Control Policies via Hybrid Control.
Jan de PriesterRicardo G. SanfeliceNathan van de WouwPublished in: ACC (2022)
Keyphrases
- control policies
- reinforcement learning
- control policy
- optimal policy
- action space
- continuous state
- function approximation
- state space
- control strategies
- markov decision processes
- control system
- finite horizon
- long run
- model free
- learning algorithm
- reinforcement learning algorithms
- temporal difference
- reward function
- multistage
- motion control
- multi agent
- optimal control
- real valued
- action selection
- robot control
- markov decision process
- mobile robot
- rl algorithms
- machine learning
- infinite horizon
- linear program
- dynamical systems