PEARL: Parallelized Expert-Assisted Reinforcement Learning for Scene Rearrangement Planning.
Hanqing WangZan WangWei LiangLap-Fai YuPublished in: CoRR (2021)
Keyphrases
- reinforcement learning
- d scene
- three dimensional
- action selection
- function approximation
- partially observable
- deterministic domains
- reinforcement learning algorithms
- video sequences
- single image
- image sequences
- markov decision processes
- model free
- multiple images
- machine learning
- blocks world
- reinforcement learning problems
- outdoor scenes
- temporal difference
- real scenes
- scene understanding
- dynamic scenes
- planning problems
- heuristic search
- optimal control
- optimal policy
- learning algorithm
- moving objects
- inverse reinforcement learning
- markov decision problems
- learning process
- function approximators
- partially observable markov decision processes
- input image
- object detection
- scene classification
- domain independent
- belief revision
- human experts
- complex scenes
- expert knowledge