Zero-Shot Policy Transfer with Disentangled Task Representation of Meta-Reinforcement Learning.
Zheng WuYichen XieWenzhao LianChanghao WangYanjiang GuoJianyu ChenStefan SchaalMasayoshi TomizukaPublished in: CoRR (2022)
Keyphrases
- reinforcement learning
- optimal policy
- policy search
- action selection
- approximate dynamic programming
- markov decision processes
- function approximation
- partially observable environments
- learning algorithm
- policy gradient
- action space
- partially observable
- meta level
- multi agent
- markov decision process
- reinforcement learning algorithms
- reinforcement learning methods
- control policies
- image representation
- state space