Planning in Learned Latent Action Spaces for Generalizable Legged Locomotion.
Tianyu LiRoberto CalandraDeepak PathakYuandong TianFranziska MeierAkshara RaiPublished in: IEEE Robotics Autom. Lett. (2021)
Keyphrases
- legged locomotion
- action space
- state space
- rough terrain
- humanoid robot
- action selection
- planning problems
- markov decision processes
- stochastic processes
- latent variables
- skill learning
- real valued
- motion planning
- random fields
- single agent
- initial state
- least squares
- markov decision problems
- heuristic search
- bayesian networks