High-speed quadrupedal locomotion by imitation-relaxation reinforcement learning.
Yongbin JinXianwei LiuYecheng ShaoHongtao WangWei YangPublished in: Nat. Mac. Intell. (2022)
Keyphrases
- reinforcement learning
- high speed
- legged robots
- robot control
- function approximation
- low power
- model free
- state space
- machine learning
- temporal difference
- iterative algorithms
- reinforcement learning algorithms
- imitation learning
- high speed networks
- probabilistic relaxation
- real robot
- multi agent
- learning algorithm
- dynamic programming
- degrees of freedom
- optimal control
- markov decision processes
- learning problems
- optimal policy
- mobile robot
- learning process
- robotic control
- rough terrain
- real time
- learning agent
- action selection
- transfer learning
- low cost
- neural network