Composable energy policies for reactive motion generation and reinforcement learning.
Julen UrainAnqi LiPuze LiuCarlo D'EramoJan PetersPublished in: Int. J. Robotics Res. (2023)
Keyphrases
- reinforcement learning
- optimal policy
- image sequences
- space time
- motion model
- human motion
- motion analysis
- markov decision process
- optical flow
- camera motion
- policy search
- motion patterns
- energy consumption
- motion capture
- control policies
- reward function
- motion estimation
- state space
- partially observable markov decision processes
- function approximation
- motion detection
- motion tracking
- reinforcement learning algorithms
- optimal control
- learning algorithm
- motion segmentation
- markov decision processes
- human body
- image segmentation
- multi agent
- control policy
- computer vision
- temporal difference
- infinite horizon
- machine learning
- humanoid robot
- spatial and temporal
- energy minimization
- motion field
- feature points
- multi agent reinforcement learning
- markov decision problems
- motion control
- multiagent reinforcement learning
- real robot
- dynamic programming
- markov random field
- model free
- agent architecture
- motion planning
- motion parameters