Assessing Policy, Loss and Planning Combinations in Reinforcement Learning using a New Modular Architecture.
Tiago Gaspar OliveiraArlindo L. OliveiraPublished in: CoRR (2022)
Keyphrases
- modular architecture
- reinforcement learning
- action selection
- reinforcement learning problems
- partially observable
- optimal policy
- markov decision problems
- partially observable markov decision processes
- reinforcement learning algorithms
- policy search
- state space
- partially observable environments
- markov decision process
- function approximators
- function approximation
- reward function
- control policies
- planning problems
- state and action spaces
- partially observable domains
- stochastic domains
- action space
- temporal difference
- dynamic programming
- reinforcement learning methods
- learning algorithm
- temporal difference learning
- deterministic domains
- partial observability
- policy gradient
- policy evaluation
- heuristic search
- classical planning
- reward shaping
- dynamical systems
- decision problems
- inverse reinforcement learning
- optimal control
- model free
- continuous state spaces
- loosely coupled
- state action
- policy iteration
- actor critic
- control policy