VER: Scaling On-Policy RL Leads to the Emergence of Navigation in Embodied Rearrangement.
Erik WijmansIrfan EssaDhruv BatraPublished in: NeurIPS (2022)
Keyphrases
- optimal policy
- reinforcement learning
- action selection
- markov decision process
- control policies
- control policy
- policy gradient
- markov decision processes
- policy search
- policy evaluation
- rl algorithms
- model free
- action space
- markov decision problems
- learning algorithm
- infinite horizon
- average reward
- decision problems
- partially observable domains
- state and action spaces
- agent learns
- reinforcement learning problems
- dynamic programming
- actor critic
- approximate dynamic programming
- mobile robot
- state space
- temporal difference
- optimal control
- indoor environments
- gradient method
- navigation systems
- autonomous vehicles