Symbolic Plans as High-Level Instructions for Reinforcement Learning.
León IllanesXi YanRodrigo Toro IcarteSheila A. McIlraithPublished in: ICAPS (2020)
Keyphrases
- high level
- reinforcement learning
- low level
- intermediate level
- function approximation
- reinforcement learning algorithms
- deterministic domains
- higher level
- planning problems
- state space
- programming language
- machine learning
- macro actions
- markov decision processes
- temporal difference
- low level features
- source code
- supervised learning
- plan generation
- learning process
- connectionist models
- plan execution
- robotic control
- state transition
- lower level
- transfer learning
- dynamic programming
- partially ordered
- markov decision process
- planning domains
- optimal control
- learning problems
- policy search
- learning algorithm