Abstraction-based branch and bound approach to Q-learning for hybrid optimal control.
Benoît LegatRaphaël M. JungersJean BouchatPublished in: L4DC (2021)
Keyphrases
- branch and bound
- optimal control
- reinforcement learning
- dynamic programming
- lower bound
- branch and bound algorithm
- search algorithm
- control problems
- optimal solution
- column generation
- combinatorial optimization
- actor critic
- upper bound
- policy iteration
- search space
- function approximation
- branch and bound procedure
- control strategy
- infinite horizon
- optimal control problems
- tree search
- feedback control
- model free
- multi agent
- state space
- branch and bound method
- randomly generated problems
- branch and bound search
- optimal policy
- reinforcement learning methods
- lagrangian heuristic
- learning rate
- learning algorithm
- genetic algorithm
- control system
- reinforcement learning algorithms
- np hard
- multi objective
- machine learning
- neural network