Alert button
Picture for Cédérick Mouliets

Cédérick Mouliets

Alert button

Reward-Based Environment States for Robot Manipulation Policy Learning

Add code
Bookmark button
Alert button
Dec 10, 2021
Cédérick Mouliets, Isabelle Ferrané, Heriberto Cuayáhuitl

Figure 1 for Reward-Based Environment States for Robot Manipulation Policy Learning
Figure 2 for Reward-Based Environment States for Robot Manipulation Policy Learning
Figure 3 for Reward-Based Environment States for Robot Manipulation Policy Learning
Figure 4 for Reward-Based Environment States for Robot Manipulation Policy Learning

Training robot manipulation policies is a challenging and open problem in robotics and artificial intelligence. In this paper we propose a novel and compact state representation based on the rewards predicted from an image-based task success classifier. Our experiments, using the Pepper robot in simulation with two deep reinforcement learning algorithms on a grab-and-lift task, reveal that our proposed state representation can achieve up to 97% task success using our best policies.

* NeurIPS Workshop on Deployable Decision Making in Embodied Systems, 2021 
Viaarxiv icon
Github IconAccess Paper or Ask Questions
Twitter IconFacebook IconLinkedin IconWhatsapp IconMessenger Icon