The Planning

1
The Planning Control of Robot Dexterous
Manipulation

• Li Han, Zexiang Li, Jeff Trinkle, Zhiqiang Qin,
  Shilong Jiang
• Dept. of Computer Science
• Texas AM University
• Dept. of Electrical and Electronic Engineering
• Hong Kong Univ. of Science and Technology

Rodin
2
Dexterous Manipulation

• Tasks a robotic hand
• grasps an object, and
• moves the object from a start configuration to a
  goal configuration.
• Assumptions
• Quasi-Static Systems
• Rigid Body Motions
• preserve distances and orientations
• Known System and Environment Parameters

3
Dexterous Manipulation Systems

• Japan

Katharina (Germany)
SAMM (O. Khatib, USA)
4
Applications
Fixture (K. Goldberg)
Digital Actor (J.-C. Latombe)
AerCam (NASA)
Cellular Man. (Sci. American)
5
Overview

• Problem Statement
• Force and Motion Feasibility Issues
• Manipulation Planning and Control
• Experimental Result
• Summary

HKUST Hand (Z. Li)
6
Dexterous Manipulation

• Feasible States
• Closure Variety or Manifold
• Feasible Velocities Tangent Vectors
• Feasible Forces Co-Tangent Vectors
• Collision-Free

7
Dexterous Manipulation

• Manipulation Planner
• Manipulation Controller
• Feasible States
• Grasp Statics Force
• Manipulation Kinematics Motion

8
Grasp Statics

• Grasp Force Feasibility and Optimization Problem

9
Grasp Statics and Friction Cones
Linear Matrix Inequality (LMI)
10
Numerical Results

• Convex Programming Involving LMIs
• (S. Boyds Convex Programming Group at Stanford)
• Feasibility and Optimization lt 7.82ms
  (HP/Convex)

11
Manipulation Kinematics

• Plan an object trajectory
• Use generalized inverse method to find a
  bestpossible joint trajectory
• Infeasible Object Trajectory?
• Contact Motion?

Manipulation Kinematics
Grasp Kinematics
12
Unreliable Manipulation Plan
13
Modular Control System Architecture
14
Experimental System Result

• Manipulation Objectives
• Move the object
• Improve the grasp

15
Future Work

• Large Scale Object Manipulation in a Crowded
  Environment
• Regrasping and Dexterous Manipulation Planning
• Dynamic Constraints
• Uncertainty and Robustness
• Applications

16
Conclusion

• Grasp Statics
• Linear Matrix Inequalities for Nonlinear Friction
  Cones
• Convex Programming
• Manipulation Kinematics
• Tangent Space (Feasibility Constraints)
• Inclusion of all kinematic variables
• A Modular Control System Architecture
• Manipulation Planning
• Local Motion in a Clear Environment