Project description

This project proposes the implementation of a cognitive agent embodied in a humanoid robot which will collaborate with a human partner or with another humanoid robot to solve a spatial problem (e.g. 3D puzzle, fold a paper). The tasks to be carried out are: (i) scene understanding: detection of objects and relations between objects, and use of these relations to infer new knowledge (i.e. reasoning); (ii) explore object affordances, learn the consequences of the actions carried out and enrich the knowledge base (i.e. learning by interaction); (iii) querying the knowledge base about what was useful in the past to predict actions that might be useful in the present, try them out and update its knowledge base (i.e. update learning). Also writing scientific papers and communicating our research advances in conferences. Methods: programming a humanoid platform using ROS2 packages, solve SLAM, use imitation learning algorithms to learn pick-and-...