Dynamic Robot Manipulation - Learned Optimization, Deformable Materials, and the Cloud
Paul G. Allen School via YouTube
Overview
Syllabus
Intro
Dynamic Robot Manipulation: Learned Optimization, Deformable Materials, and the Cloud Jeffrey Ichnowski, Ph.D.
Rigid World
Combine Grasp, Motion, and Dynamics
Planning for Dynamic Deformable
Use the Cloud
Trajectory and Optimization Discretization
Obstacle Constraint Linearization
Time Optimization
Science Robotics
Quadratic Programs in Robotics
OSQP Algorithm Overview
RLQP Reinforcement Learning QP Solver
Experiment Problem Classes Randomly generated QPs from OSQP benchmark suite
Benchmark Problems and Generalization
Analysis of a Learned Policy
Introducing Inertial Constraints
Suction Transport
Suction Constraint Analytic model
Suction Failures and Deformation
Suction Cup Deformation Constraint
Suction Gripper w/ Embedded Sensor
Learning a Suction Constraint
Suction Constraint Learning Pipeline
Robots of the Lost Arc Tasks
Problem Setup
Trajectory Parameterization
Sensitivity to Initial Conditions
Reset Motion
INDy: Self-Supervised Training
Cloud-based 72-core Motion Planning
Motion Planning Compute Requirements
Serverless Computing Limitations
Serverless Computing Environment
Taught by
Paul G. Allen School