Reinforcement Learning with Sparse Bellman Error Extrapolation for Infinite-Horizon Approximate Optimal Regulation.
Max L. GreenePatryk DeptulaScott NivisonWarren E. DixonPublished in: CDC (2019)
Keyphrases
- infinite horizon
- optimal control
- reinforcement learning
- dynamic programming
- optimal policy
- finite horizon
- markov decision processes
- average cost
- partially observable
- stochastic demand
- piecewise linear
- single item
- state space
- long run
- markov decision process
- total reward
- fixed cost
- policy iteration
- production planning
- expected cost
- holding cost
- inventory policy
- asymptotically optimal
- actor critic
- production capacity
- linear program
- reinforcement learning algorithms
- average reward
- learning algorithm
- periodic review
- function approximation
- finite state
- optimal solution
- inventory models
- multistage
- control strategy
- search space
- state dependent
- state action
- inventory control
- multi agent
- sufficient conditions
- single product
- model free