SPOTTER: Extending Symbolic Planning Operators through Targeted Reinforcement Learning.
Vasanth SarathyDaniel KasenbergShivam GoelJivko SinapovMatthias ScheutzPublished in: CoRR (2020)
Keyphrases
- reinforcement learning
- action selection
- function approximation
- stochastic domains
- state space
- heuristic search
- deterministic domains
- macro actions
- planning problems
- partially observable
- automated planning
- mixed initiative
- high level
- reward shaping
- multi agent systems
- optimal control
- markov decision processes
- markov decision problems
- ai planning
- domain independent
- macro operators
- reinforcement learning problems
- partial observability
- partially observable domains
- optimal policy
- multi agent reinforcement learning
- blocks world
- complex domains
- symbolic representation
- goal oriented