Evolving Robust Policy Coverage Sets in Multi-Objective Markov Decision Processes Through Intrinsically Motivated Self-Play.
Sherif M. AbdelfattahKathryn KasmarikJiankun HuPublished in: Frontiers Neurorobotics (2018)
Keyphrases
- markov decision processes
- optimal policy
- multi objective
- policy iteration
- markov decision process
- finite horizon
- average reward
- average cost
- infinite horizon
- state space
- state and action spaces
- partially observable
- finite state
- decision processes
- action space
- reward function
- reinforcement learning
- dynamic programming
- transition matrices
- reachability analysis
- multi objective optimization
- evolutionary algorithm
- planning under uncertainty
- policy evaluation
- markov decision problems
- decision theoretic planning
- factored mdps
- discounted reward
- total reward
- decision problems
- policy iteration algorithm
- objective function
- expected reward
- action sets
- stationary policies
- reinforcement learning algorithms
- initial state
- partially observable markov decision processes
- long run
- multistage
- action selection
- model based reinforcement learning
- optimal solution
- continuous state spaces
- optimal control
- sufficient conditions
- semi markov decision processes
- search space