SPOTTER: Extending Symbolic Planning Operators through Targeted Reinforcement Learning.
Vasanth SarathyDaniel KasenbergShivam GoelJivko SinapovMatthias ScheutzPublished in: AAMAS (2021)
Keyphrases
- reinforcement learning
- action selection
- partially observable
- function approximation
- macro actions
- deterministic domains
- goal oriented
- macro operators
- high level
- heuristic search
- planning problems
- motion planning
- multi agent
- state space
- ai planning
- partially observable markov decision processes
- machine learning
- single agent
- mixed initiative
- reinforcement learning problems
- planning systems
- planning process
- blocks world
- partial observability
- reinforcement learning algorithms
- temporal difference
- model free
- domain independent
- transfer learning
- learning algorithm
- complex domains
- plan generation
- optimal policy
- decision support
- dynamic programming
- learning process
- neural network
- reward shaping