Challenges for the policy representation when applying reinforcement learning in robotics.
Petar KormushevSylvain CalinonDarwin G. CaldwellBarkan UgurluPublished in: IJCNN (2012)
Keyphrases
- reinforcement learning
- optimal policy
- markov decision process
- computer vision
- policy search
- markov decision processes
- machine learning
- robot control
- partially observable
- representation scheme
- action selection
- open issues
- control policies
- state space
- reinforcement learning problems
- policy gradient
- transition model
- state and action spaces
- action space
- policy iteration
- optimal control
- key issues
- lessons learned
- dynamic programming
- multi agent
- artificial intelligence
- learning algorithm