MetaDrive: Composing Diverse Driving Scenarios for Generalizable Reinforcement Learning.
Quanyi LiZhenghao PengZhenghai XueQihang ZhangBolei ZhouPublished in: CoRR (2021)
Keyphrases
- reinforcement learning
- real world
- function approximation
- reinforcement learning algorithms
- state space
- multi agent
- wide variety
- learning process
- sufficient conditions
- temporal difference
- model free
- optimal control
- optimal policy
- dynamic programming
- policy search
- stochastic approximation
- temporal difference learning
- markov decision processes
- transition model
- relational reinforcement learning
- application scenarios
- learning scenarios
- transfer learning
- supervised learning
- information systems
- computer vision