Learning to detect the functional components of doorbell buttons using active exploration and multimodal correlation

摘要: This paper describes a large-scale experimental study, in which humanoid robot learned to press and detect doorbell buttons autonomously. The models for action selection visual detection were grounded the robot's sensorimotor experience without human intervention. Experiments performed with seven buttons, provided auditory feedback when pressed. predict locations of functional components each button accurately. trained model was also able novel buttons.