Post on 11-Feb-2017
Kensuke Harada∗1, Tokuo Tsuji∗2, Kohei Kikuchi∗3
Kazuyuki Nagata∗1, Hiromu Onda∗1, Yoshihiro Kawai∗1
∗1 National Inst. of AIST∗2 Kyushu Univ.
∗3 Toyota Motors Co. Ltd.
National Institute ofAdvanced Industrial Science and Technology (AIST)
Base Position Planning for Dual-arm Mobile ManipulatorsPerforming a Sequence of Pick-and-place Tasks
INTRODUCTIONParts Supply Task
Collect necessary parts from the storage area
Determine a sequence of the base position
Selective Use of Two Hands
Can Reduce the Number of Sequence
Dual-arm Mobile ManipulatorPerforming Parts-Supply Task
OVERVIEW OF PROPOSED METHOD
Dual-arm mobile manipulator collects one part from each box.
Shipping Carton
Example
OVERVIEW OF PROPOSED METHOD
Region on base position where right hand can grasp an object in the box 1
Region on base position where left hand can grasp an object in the box 1
OVERVIEW OF PROPOSED METHOD
Obtain region on base position for all the boxes
OVERVIEW OF PROPOSED METHOD
Obtain a sequence of base position
OVERVIEW OF PROPOSED METHODMinimize : Number of Sequence and Travel Distance
Common Area of Base position: ・ Left Hand Picks an Object in the Box 1・ Right Hand Pick an Object in the Box 2
Minimizing the travel distance
FEATURES OF PROPOSED METHOD
• Objects stored in a box is known• The number and the pose are not known
• Stable grasp • Collision avoidance when grasping
• Minimizing the number of sequence• Minimizing the travel distance
Base Position Planning for Dual-arm Mobile ManipulatorsPerforming a Sequence of Pick-and-place Tasks
RELATED WORKS
Base Position for Mobile Manipulator Maximizing the Manipulability F.G.Pin et al. (1990), H. Seragi et al. (1993), K. Tchon et al. (2000) 等
Base Position Considering the Stable GraspR. Diankov et al. (2008), N. Vahrenkamp et al. (2012)
Manipulation Planner for Dual-arm ManipulatorY. Koga et al. (1994), J.-P. Saut et al. (2010), K. Harada et al. (2012)
IN THIS PRESENTATION
1. Inverse Kinematics Considering the Stable Grasp
2. Region on Mobile Base
3. Formulation of Optimization Problem
4. Numerical Examples
INVERSE KINEMATICS CONSIDERING THE STABLE GRASP ( 1/3 )
Collision Avoidance
Joint Limit
Ellipsoidal Approximation of Links and Environments
Using EQuadProg++
QP Problem to Obtain Joint Velocity
INVERSE KINEMATICS CONSIDERING THE STABLE GRASP ( 2/3 )Grasping Posture Database
• Quadratic Approximation of Object Surface
• Grasp Planning
• Construction of Grasping Posture Database
INVERSE KINEMATICS CONSIDERING THE STABLE GRASP ( 3/3 )
1. Connect Initial Pose and 1st Data using Straight Line Segment
2. Split Segment into Small Steps
3. Solve QP Problem for Each Step
4. If QP problem does not have solution, proceed to next data
Search for the Grasping Posture Database
IN THIS PRESENTATION
1. Inverse Kinematics Considering the Stable Grasp
2. Region on Mobile Base
3. Formulation of Optimization Problem
4. Numerical Examples
REGION ON MOBILE BASE
Object pose is not known
Mobile Base Pose Stably Grasping of an Object
Discretize the Box / Moving Area
REGION ON MOBILE BASESearching Algorithm
STEP1Base Pose Stably Grasping All Objects in the Box
Using Multi-thread
REGION ON MOBILE BASESearching Algorithm
STEP1
STEP2Checking Neighboring Points
Base Pose Stably Grasping All Objects in the Box
REGION ON MOBILE BASE
STEP1
STEP2Checking Neighboring Points
Base Pose Stably Grasping All Objects in the Box
Searching Algorithm
REGION ON MOBILE BASESearching Algorithm
STEP1
STEP2Checking Neighboring Points
Base Pose Stably Grasping All Objects in the Box
REGION ON MOBILE BASESearching Algorithm
STEP1
STEP2Checking Neighboring Points
Base Pose Stably Grasping All Objects in the Box
IN THIS PRESENTATION
1. Inverse Kinematics Considering the Stable Grasp
2. Region on Mobile Base
3. Formulation of Optimization Problem
4. Numerical Examples
DEFINITION
P1 P2 P3 P4 P5 P6
P = [P1 P2 P3 P4 P5 P6]P2P1∩ P3P2∩ P4P3∩ P6P5∩
s=[ 1, 2, 3, 4, 5, 6, (1,2), (2,3), (3,4), (5,6)] =[ s1, s2, s3, s4, s5, s6, s7, s8, s9, s10]
FORMULATION
Which Area to Visit 0-1 Knapsack Problem
Minimizing: Number of Sequence Branch and Bound (BB) Method
Number of Sequence
FORMULATION
Order of Area to Visit
Minimizing: Traveling Distance Traveling Salesman Problem
Simulated Annealing
EXAMPLES
Discretizing Box Area
Region on Mobile Base Realizing the Stable Grasp
Region (Grasped by Left Hand)
Nextage + Mobile Base
Region (Grasped by Right Hand)
RESULTSDistance between Two Boxes : 0
1st Base Position : Pick an Object from Boxes 1, 2, 3 and 4
RESULTS
1st Base Position : Pick an Object from Boxes 1 and 22nd Base Position : Pick an Object from Boxes 3 and 4
Distance between Two Boxes : 0.4 [m]
RESULTSDistance between Two Boxes : 0.8 [m]
1st Base Position : Pick an Object from Box 12nd Base Position : Pick an Object from Box 23rd Base Position : Pick an Object from Box 34th Base Position : Pick an Object from Box 4
EXAMPLES
1st Base Position
2nd Base Position
Initial Pose
CALCULATION TIME
This method is to be used for offline calculation of base position
SUMMARY• A sequence of the base position is planned for dual-arm
mobile manipulators performing multiple pick-and-place tasks.
• A robot can selectively use either the right or the left hand to pick up an object and can minimize the sequence size of the base position.
• A region on the base position is obtained where a robot
can pick up an object placed anywhere in a box. • A sequence of base positions is obtained by combining the
branch and bound and simulated annealing methods.