2018/09/27 by Jonathan Juett, Benjamin Kuipers, Juett, Jonathan +1
Medicine · Psychology · #Action Observation and Synchronization #Artificial Intelligence (cs.AI) #Cerebral Palsy and Movement Disorders #Child and Animal Learning Development #FOS: Computer and information sciences #Machine Learning (cs.LG) #Robotics (cs.RO)
paper · pdf · doi:10.48550/arxiv.1809.10788
openalex publication_date 2018/09/27 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
The young infant explores its body, its sensorimotor system, and the\nimmediately accessible parts of its environment, over the course of a few\nmonths creating a model of peripersonal space useful for reaching and grasping\nobjects around it. Drawing on constraints from the empirical literature on\ninfant behavior, we present a preliminary computational model of this learning\nprocess, implemented and evaluated on a physical robot. The learning agent\nexplores the relationship between the configuration space of the arm, sensing\njoint angles through proprioception, and its visual perceptions of the hand and\ngrippers. The resulting knowledge is represented as the peripersonal space\n(PPS) graph, where nodes represent states of the arm, edges represent safe\nmovements, and paths represent safe trajectories from one pose to another. In\nour model, the learning process is driven by intrinsic motivation. When\nrepeatedly performing an action, the agent learns the typical result, but also\ndetects unusual outcomes, and is motivated to learn how to make those unusual\nresults reliable. Arm motions typically leave the static background unchanged,\nbut occasionally bump an object, changing its static position. The reach action\nis learned as a reliable way to bump and move an object in the environment.\nSimilarly, once a reliable reach action is learned, it typically makes a\nquasi-static change in the environment, moving an object from one static\nposition to another. The unusual outcome is that the object is accidentally\ngrasped (thanks to the innate Palmar reflex), and thereafter moves dynamically\nwith the hand. Learning to make grasps reliable is more complex than for\nreaches, but we demonstrate significant progress. Our current results are steps\ntoward autonomous sensorimotor learning of motion, reaching, and grasping in\nperipersonal space, based on unguided exploration and intrinsic motivation.\n