Date of this Version
We propose that some aspects of task based learning in robotics can be approached using nativist and constructivist views on human sensorimotor development as a metaphor. We use findings in developmental psychology, neurophysiology, and machine perception to guide a robotic learning system's level of representation both for actions and for percepts. Visually driven grasping is chosen as the experimental task since it has general applicability and it has been extensively researched from several perspectives. An implementation of a robotic system with a dexterous three fingered hand, compliant instrumented wrist, arm and vision is used to test these ideas. Several sensorimotor primitives (vision segmentation and manipulatory reflexes) are implemented in this system and may be thought of as the "innate" perceptual and motor abilities of the system.
Applying empirical learning techniques to real situations brings up some important issues such as observation sparsity in high dimensional spaces, arbitrary underlying functional forms of the reinforcement distribution and robustness to noise in exemplars. The well established technique of non-parametric projection pursuit regression (PPR) is used to accomplish reinforcement learning by searching for generalization directions determining projections of high dimensional data sets which capture task invariants. Additionally, the learning process generally implies failures along the way. Therefore, the mechanics of the untrained robotic system must be able to tolerate grave mistakes during learning and not damage itself. We address this by the use of an instrumented compliant robot wrist which controls impact forces.
Marcos Salganicoff and Ruzena Bajcsy, "Robotic Sensorimotor Learning in Continuous Domains", . September 1991.
Date Posted: 14 August 2007