Metric-based imitation learning in humans and robots

点击翻译按钮获取中文摘要

Learning by imitation is a versatile and rapid mechanisms to transfer motor skills from one intelligent agent (humans, animals and robots) to another – which can be observed in Nature as well as applied in form of “programming by demonstration” in artificial systems. Infants, for example, react to the perception of facial gestures by producing similar behavior. In robotics, computational approaches to motor learning by imitation are long considered the most promising path to success. The development of autonomous robotic systems that can learn from human demonstrations to imitate a desired behavior - rather than being manually programmed - has huge technological potential with applications in manufacturing, elderly care and the service industry and is currently at the focus of robotics research. Unlike trial-and-error-based learning methods such as reinforcement learning, imitation allows rapid learning. Approaches to imitation learning have delivered huge success ranging from helicopter acrobatics, high-speed arm skills, haptic control, gestures, manipulation to legged locomotion. The machine learning algorithms that make imitation learning possible are well studied.Surprisingly, despite all of these impressive successes in the acquisition of new motor skills in robotic systems by imitation learning, fundamental scientific research questions in imitation learning of central importance have remained open for decades. Among such core questions is the one of the correspondence problem: how can one agent (the learner or imitator) produce a similar behavior - in some aspect - with behavior it perceives in another agent (the expert or demonstrator) given that the two agents have different kinematic and dynamics (body morphology, degrees of freedom, constraints, joints and actuators, torque limits), or in other words, cover different state spaces?Thus, the goal of this project is to use a metric understanding of embodiment to improve robotic motor skills through expert observations. We aim to shed light into important fundamental research questions on the (i) role of learner’s embodiment in statistical imitation learning, (ii) how the correspondence problem can be formalized properly, (iii) how can the behavior transferability vs task complexity dilemma be resolved and (iv) how to develop new statistical deep imitation learning algorithms based on these insights.