VUB, Artificial Intelligence Lab
Action Learning and Grounding in Simulated Human-Robot Interactions
Proposal of a novel framework for manipulation action learning and grounding of shape, color, and action words through reinforcement and cross-situational learning. The latter is also used to identify auxiliary words and phrases. The proposed framework is evaluated through simulated human-robot interactions. The results show that the framework enables the robot to successfully learn manipulation of different objects and to ground all employed words through their corresponding percepts.
Oliver Roesler is a PhD student at the Brain Embodiment Lab at the University of Reading and a researcher at the Artificial Intelligence Lab at the Vrije Universiteit Brussel working on a framework to automatically connect semantic and perceptual information to enable more natural and complex human-robot interactions.