Transferring Human Skills to Humanoid Robots Dongheui Lee dhlee@tum.de Dynamic Human-Robot-Interaction for Automation Systems (HRI) Lab b Department of Electrical Engineering and Information Technology Technical University of Munich Technical University of Munich
Transferring Human Skills to Humanoid Robots Movements Manipulation Pysical HRI • learning • learning • whole body • whole body • contact • contact motion coordination estabilishment • recognition • grasping skills • physical • reproduction • reproduction • interaction • interaction coaching coaching force control • haptic assistance policy in collaboration
Programming by Demonstration Mirror Neurons Mirror Neurons Observing man picking feed Monkey picking feed Activities of Mirror Neuron （ F5 ） Monkey Brain: F5 [Gallese et al.1996] [Rizzolatti et al. 1996].
Programming by Demonstration Mathematical formulation of Mirror Neurons Mathematical formulation of Mirror Neurons • Mimesis Model • Probabilistic representation for Probabilistic representation for spatiotemporal data • Learning, recognition, Learning, recognition, generation (a bidirectional a 11 a 22 a N-1 N-1 a NN computational model) a a 12 a a N-1 N … S 2 S 1 S N S N-1 • Mimesis from partial observation a N1 b 1 b 1 b 2 b 2 b N 1 b N-1 b N b N [Lee and Nakamura IJRR 2010]
Programming by Demonstration Mathematical formulation of Mirror Neurons Mathematical formulation of Mirror Neurons • Mimesis Model • Probabilistic representation for Probabilistic representation for spatiotemporal data • Learning, recognition, Learning, recognition, generation (a bidirectional computational model) • Mimesis from partial observation [Lee and Nakamura IJRR 2010] [Lee and Nakamura IJRR 2010]
Motion Recostruction from Monocular Vision
Recognition from Optical Flow Biological Movement [Johansson 1975] � Aim to recognize and recover the motion from the optical flow of feature points
{C} {I} {D} { 2D Optical Flow p C T D T λ λ λ I λ λ C → D → → θ θ x x x x x Camera Motion Primitives Image Demonstrator Cartesian in Joint Space { Joint Cartesian Cartesian angle, angular velocity, HMM HMM base-joint velocity} Human Humanoid behavior behavior Generation Recognition 3D Whole 3D Whole Partial Observation: Conditioned from partial body Motion by Observation observation Occluded Monocular Image
Motion Recostruction from Monocular Vision [Lee and Nakamura IROS 2007] [Lee and Nakamura IROS 2007] Human perception of biological movements Human perception of biological movements magenta : True Model • Activity recognition blue: Recovered Model - 6 motions • Motion Capturing 56DOF
Transferring Human Skills to Humanoid Robots Movements Manipulation Pysical HRI • learning • learning • full body • full body • contact • contact motion coordination estabilishment • recognition • grasping skills • physical • reproduction • reproduction • interaction • interaction coaching coaching force control • haptic assistance policy in collaboration
Grasping Skill Learning from Motion and Force Data • Learning grasping skills from motion and force patterns • Teleoperation using Cyberglove, Flock of Birds, & C b Cybergrasp (Haptic Feedback) (H ti F db k)
Grasping Skill Reproduction • Parallel position (PD) and force (PI) control + + = τ − & & & & T ( ) ( , ) ( ) M q q C q q q g q J f ∫ ∫ τ τ = = + + + + + + & T { { } } k k e e k k e e J J f f k k e e dt dt p p d d p d d f f f f z x
Learning Interaction vs. Internal Forces Interaction force learned Static internal force learne Static internal force learne demonstrations
Generalization Capability: Radius
Transferring other manipulation skills Mechanism for Association of Whole Body Motion from � Tool Knowledge Tool in Body Schema [Maravita and Iriki 2004] Tool in Body Schema [Maravita and Iriki 2004] � e.g. Distal-type neurons � [Lee et al IROS2008] [Kunori,Lee,NakamuraIROS2009] � Learning interaction control policies � Dynamic movement primitives for parallel position and � force control Deformable objects, sculpting tasks j , p g � [Koropouli, Lee, Hirche, 2011 IROS] �
Transferring Human Skills to Humanoid Robots Movements Manipulation Pysical HRI • learning • learning • whole body • whole body • contact • contact motion coordination estabilishment • recognition • grasping skills • physical • reproduction • reproduction • interaction • interaction coaching coaching force control • haptic assistance policy in collaboration
Simple Human Robot Interaction AUTOMATICA 2010 Collaboration with Dr Ott Dr Albu-Schaeffer Haddadin DLR Collaboration with Dr. Ott, Dr. Albu Schaeffer, Haddadin, DLR
Imitation Learning Imitation Learning Execution Teaching g Physical interaction Mimetic Communication [Lee et al IJRR 2010]
Motivation: Motion � Interaction From the Movie “Terminator 2 Judgment Day” Issues for pHRI: � H � Human motion imitation � Marker Control � M ti i it ti k C t l � Learn/Recognize/Generate Motion Primitives � Mimesis Model � Learn/Recognize/Generate Interaction Rules � Mimetic Communication Model ea / ecog e/Ge e ate te act o u es et c Co u cat o ode � Contact transition � Real-time motion adaptation � Application : High-Five like interaction
Motion Imitation by Marker Control Dynamics of the humanoid’s upper body on a free-floating base body: τ ⎛ ⎞ ⎛ ⎞ ⎛ ⎞ ⎛ ⎞ ⎛ ⎛ ⎞ ⎞ && && & & q q + = & & ( ) ( , , ) ⎜ ⎟ ⎜ ⎟ ⎜ ⎟ M q C q q x && & ⎝ ⎠ ⎝ ⎠ ⎝ ⎠ x x f Virtual Springs: 1 2 = − ( ( , ) ) ( ( , ) ) V q x V q x k r k r r q x r q x , i i d i i 2 Marker pos. of the Measured marker simulation position τ ⎛ ⎞ ⎛ ⎞ & q ⎝ ⎠ ∑ = − + − T ⎜ ⎟ ( ) ⎜ ⎟ ( )( ( , )) D q k J q r r q x i i d i , i & ⎝ ⎠ f x ∀ ∈ i M
Motion Imitation by Marker Control • Upper body Control: Marker trajectory following • Lower body Control: Balancing, Hip orientation and Height following [Ott, Lee, Nakamura, “Motion Capture based Human Motion Recognition and Imitation by Direct Marker Control”, Humanoids 2008]
Full Body Motion Imitation
Motion Learning � Interaction Learning interactive interactive HMM HMM primitive primitive motion motion HMM HMM primitive HMM HMM HMM HMM motion motion recognition generation primitive primitive behavior behavior behavior behavior behavior human human robot robot Mimetic Communication Model l learning, recognition & generation i i i & i of interaction primitives • How to react to human’s action • Contact location & timing
Physical Contact Establishment • Adaptation of the robot’s motion to the desired contact point in real-time: 1) Use additional spring (red) connected to the desired contact point. 2) Project the forces of the hand’s marker springs (green) into a subspace related to the hand orientation. i b l d h h d i i τ ⎛ ⎞ ⎛ ⎞ & q ∑ = − + − T ⎜ ⎟ ( ) ⎜ ⎟ ( )( ( , )) D q k J q r r q x , i i d i i & ⎝ ⎝ ⎠ ⎠ ⎝ ⎠ ⎝ ⎠ f x ∀ ∈ \ i M H δ + − δ ⎛ ⎞ (1 ) F F ∑ + , , k h k k w k T ⎜ ⎟ ( ) J q h k , T ⎝ ⎠ = , k R L w k , Distance information � smooth transition contact/non-contact • Position control � (Position based) Impedance control • Position control � (Position based) Impedance control � Limiting the contact forces � Implementing “smooth” contact
Experiments • 12 motion primitives and 8 interaction primitives • Implementation to humanoid robot (38DOF), 30DOF is controlled. • Position based Impedance Control to the Upper body p pp y [Lee, Ott, Nakamura, ICRA 2009] [Lee, Ott, Nakamura, IJRR 2010]
Imitation Learning Imitation Learning Execution Teaching g Physical interaction Mimetic Communication Physical coaching [Lee et al IJRR 2010] [Lee & Ott, Autonomous Robots 2011] [ ]
Demonstration Technique Observational Demo Observational Demo Kinesthetic Demo Kinesthetic Demo synchronized whole body • Unsynchronized body motion motion • Accidental disturbance correspondence problem No correspondence problem
Overview
Recommend
More recommend
