MetaDrive: Composing Diverse Driving Scenarios for Generalizable Reinforcement Learning.
Quanyi LiZhenghao PengLan FengQihang ZhangZhenghai XueBolei ZhouPublished in: IEEE Trans. Pattern Anal. Mach. Intell. (2023)
Keyphrases
- reinforcement learning
- real world
- state space
- model free
- machine learning
- reinforcement learning algorithms
- wide variety
- learning algorithm
- multi agent
- markov decision processes
- optimal policy
- decision making
- neural network
- dynamic programming
- function approximation
- temporal difference
- rich set
- robotic control
- multi agent reinforcement learning
- relational reinforcement learning
- real time
- optimal control
- sufficient conditions
- search space
- information retrieval
- databases
- data sets