Robotics and Animatronics in Disney 2002-03 Postdoctoral Fellow, Carnegie Mellon University Lecture 1: Overview (Prof
Total Page:16
File Type:pdf, Size:1020Kb
7/21/2013 About Myself 2002 Ph.D. in Mechanical Engineering, University of Tokyo (Department of Mechano-Informatics) Robotics and Animatronics in Disney 2002-03 Postdoctoral Fellow, Carnegie Mellon University Lecture 1: Overview (Prof. Jessica Hodgins) 2003-08 Assistant/Associate Professor, Department of Katsu Yamane Mechano-Informatics, University of Tokyo 2008- Senior Research Scientist, Disney Research, [email protected] Pittsburgh Research Interests The Walt Disney Company Synthesizing natural motions for humanoid robots and • Studio Entertainment: Walt Disney Animation Studios, Pixar, characters Walt Disney Studios Motion Pictures, Marvel, (Lucasfilm), etc. • Tools: dynamics simulation, humanoid robot modeling, • Parks and Resorts: Walt Disney Imagineering, Walt Disney inverse kinematics, geometry-based algorithms World, Disney Cruise Line, etc. • Analysis: motion capture, musculoskeletal human model • Media Networks: ABC, ESPN, ABC Family, etc. • Synthesis: humanoid robot control, using motion capture • Disney Consumer Products: Disney Consumer Products, data, motion planning Disney Publishing Worldwide, Disney Store • Disney Interactive: Disney Interactive Games, Disney Interactive Media Disney Research DRP • A part of Walt Disney Imagineering • Director: Jessica Hodgins (also LA) • But works with all business units • 4 [Senior] Research Scientists (1 in robotics) • Locations • 8 postdocs (3 in robotics) – Disney Research, LA (DRLA): Glendale, CA • 5 research staffs – Disney Research, Pittsburgh (DRP): Pittsburgh, PA • 4 administrative staffs – Disney Research Boston (DRB): Boston, MA • Consultants – Disney Research, Zurich (DRZ): Zurich, Switzerland • Many interns – Pixar Research Group: Emeryville, CA – Walt Disney Animation Studios Research: Burbank, CA 1 7/21/2013 DRP Research Areas Robotics @ Disney • Computer vision • Studio Entertainment: Walt Disney Animation Studios, Pixar, • Human-computer interaction Walt Disney Studios Motion Pictures, Marvel, (Lucasfilm), etc. • Robotics • Parks and Resorts: Walt Disney Imagineering, Walt Disney World, Disney Cruise Line, etc. • Machine learning • Media Networks: ABC, ESPN, ABC Family, etc. • Wireless communication • Disney Consumer Products: Disney Consumer Products, • Speech recognition Disney Publishing Worldwide, Disney Store • Disney Interactive: Disney Interactive Games, Disney Interactive Media Robotics @ Disney Robotics @ Disney • Walt Disney Imagineering • Just accomplishing a task is not enough: style and – Mostly development of robots for deployment plausibility (and cost) are also important. – Advanced development: semi-long-term projects • We have existing hardware systems. • Disney Research • Most “customers” don’t care technology. – Pittsburgh: 1 SRS, 3 postdocs, 4-6 lab associates (interns) – Artists – LA: 2 SRS, 1 postdoc – Engineers – Zurich: 2 robotics-related projects – Guests – Consumers • We can usually control the environment. – We may even cheat to some extent. Robotics @ Disney Natural and Stylized Motions • Animation – Highly labor-intensive task – Experienced animators – Physical robot for fine-tuning • Maintenance – Manual adjustments – Limited time/resource Natural and Humanoid Human-Robot Stylized Modeling and Interaction Motions Control 2 7/21/2013 Humanoid Human-Robot Modeling and Interaction Control • Disney robots (Audio-Animatronic Figures) • Some prior attempts in Disney – No real contacts with environment / objects • Operator + supervisor: not scalable Challenges DRP Projects Human-Robot Interaction Safe physical Animation: interfaces Robust force control Soft robots interaction that are simple but Physical human-robot interaction allows full control Modeling and sensing New hardware Mocap to non-humanoids Maintenance: maintain Simple and inherently Leverage existing original animation robust robots A100 modeling & control Locomotion on dynamic objects hardware quality Programming interfaces for animators Contact force optimization Mocap to humanoids Natural and Humanoid Natural and Humanoid Human-Robot Stylized Modeling and Stylized Simplified models Modeling and Interaction Motions Control Motions Identification & calibration Control Robot Platforms “Sky” • Typical Disney figure (A100) • Hydraulic (600PSI) • 4DOF body, 3DOF torso, 2 shoulder shrugs (linear), 7DOF arms, 3DOF neck, mouth, 10 fingers • Disney proprietary low-level position controller • Linux interface to update position commands and obtain position / pressure data @ 120Hz “Sky” (A100 figure) @ DRP “The Sarcos Robot” @ CMU “Miley” @ DRLA 3 7/21/2013 The Sarcos Robot “Miley” • Floating-base humanoid robot • Planar bipedal robot developed by Rick Cory (DRLA) • Hydraulic (3000PSI) • 2 joints per leg • 7DOF legs, 7DOF arms, 3DOF torso, 3DOF neck • Series elastic actuators • Sensors: joint angles (potentiometer), joint torque, IMU, 6-axis force/torque sensors in feet • On-board low-level position/force controller @ 5kHz • Xenomai-based external PC for high-level control @ 1kHz Day 1 • Introduction • Computation Lecture Overview – Inverse and Forward Dynamics – Geometric Algorithms for Robotics a bk f Day 2 Day 3 • Motion Synthesis • Modeling and Control – Review of physics-based techniques in animation – Humanoid robot modeling and control – Using motion capture to control humanoid robots – Human modeling and control – Using motion capture to control non-humanoid robots • Physical human-robot interaction 4 7/21/2013 Evaluation • Attendance • Report: an essay on the discussion points presented at one of the 8 lectures (details to follow) 5 .