CSCI498B/598B Human-Centered Robotics Nov 05, 2014 Slides of LfD are adapted from Dr. Aude G. Billard Gesture Recognition Biological Inspiration Robotic Learning by Imitation Implementation Motor Learning Gesture Recognition Robotic Learning by Imitation Implementation Motor Learning Imitation Learning in Robots Granularity Cognition How to imitate? Level 3: Learning primitives of motion Level 2: Reproduction of trajectories Level 1: One-shot learning Level 0: Implicit imitation Learning What to imitate The robot should learn that the important feature in this task is that the queen should be moved 2 steps forward vertically Once the robot has learned the rule of motion for the queen, it can apply this rule for moving the queen from locations not seen during the demonstrations Transmitting human skills and knowledge to robots Learning a Packaging Task IMITATION LEARNING VERSUS MOTOR LEARNING Imitation learning – Programming by Demonstration: • A way to speed up learning, to reduce the search space • A way to share with the robot’s the same vocabulary of motor skills Self Motor Learning - Reinforcement Learning • To adapt to novel situations • To adapt the demonstrated motions to the robot’s body IMITATION LEARNING VERSUS MOTOR LEARNING Imitation learning – Programming by Demonstration: • A way to speed up learning, to reduce the search space • A way to share with the robot’s the same vocabulary of motor skills Self Motor Learning - Reinforcement Learning • To adapt to novel situations • To adapt the demonstrated motions to the robot’s body From Recognizing to Reproducing Gestures A.G. Billard - SHS Program in Cognitive Psychology - Spring 2007 From Recognizing to Reproducing Gestures Recovers generalized signal by regression GMM Encoding: Mixture of k Gaussians A.G. Billard - SHS Program in Cognitive Psychology - Spring 2007 From Recognizing to Reproducing Gestures Gaussian mixtures Gaussian mixture model (GMM) Gaussian mixture regression (GMR) 12 From Recognizing to Reproducing Gestures Gaussian mixtures 13 From Recognizing to Reproducing Gestures Calion, S., Guenter, F. and Billard, A. (2007) On Learning, Representing and Generalizing a Task in a Humanoid Robot. IEEE A.G.37:2. Billard - Part SHS Program in Cognitive Psychology - Spring 2007 Transactions on Systems, Man and Cybernetics, B. Special issue on robot learning by observation, demonstration and imitation. IMITATION LEARNING VERSUS MOTOR LEARNING Imitation learning – Programming by Demonstration: • A way to speed up learning, to reduce the search space • A way to share with the robot’s the same vocabulary of motor skills Self Motor Learning - Reinforcement Learning • To adapt to novel situations • To adapt the demonstrated motions to the robot’s body IMITATION LEARNING VERSUS MOTOR LEARNING Imitation learning – Programming by Demonstration: • A way to speed up learning, to reduce the search space • A way to share with the robot’s the same vocabulary of motor skills Self Motor Learning - Reinforcement Learning • To adapt to novel situations • To adapt the demonstrated motions to the robot’s body Adapting the Demonstration to Fit the Robot’s Body A.G. Billard - SHS Program in Cognitive Psychology - Spring 2007 Adapting the Demonstration to Fit the Robot’s Body A.G. Billard - SHS Program in Cognitive Psychology - Spring 2007 Dynamic Adaptation of Gesture Reproduction Dynamical system modulation to be robust to perturbations A.G. Billard - SHS Program in Cognitive Psychology - Spring 2007 Dynamical System Modulation Different initial conditions Adaptation to sudden target displacement Different initial conditions Adaptation to sudden target displacement Limitation of the System If the novel situation differs Importantly from the demonstrated one, then adapting the demonstrated trajectory is no longer sufficient to satisfy the task. Limitation of the System If the novel situation differs Importantly from the demonstrated one, then adapting the demonstrated trajectory is no longer sufficient to satisfy the task. Need to relearn the task -- Reinforcement Learning Need to define a new metric – the reward RL - To Adapt to Novel Situations A.G. Billard - SHS Program in Cognitive Psychology - Spring 2007 RL - To Adapt to Novel Situations SUMMARY Learning new tasks relies on various means of teaching the robots. Imitation learning is useful in that it gives hints as to the optimal solution The robot must however rely on generic skills of its own to adapt the demonstration to its own body and to the context Learning of complex skills is overall relatively slow and must proceed incrementally 27 28 29 30