TOPOLOGICAL COMPLEXITY OF KINEMATIC MAPS

Slides:



Advertisements
Similar presentations
COMP Robotics: An Introduction
Advertisements

Outline: Introduction Link Description Link-Connection Description
1 C02,C03 – ,27,29 Advanced Robotics for Autonomous Manipulation Department of Mechanical EngineeringME 696 – Advanced Topics in Mechanical Engineering.
Feasible trajectories for mobile robots with kinematic and environment constraints Paper by Jean-Paul Laumond I am Henrik Tidefelt.
Kinematics & Grasping Need to know: Representing mechanism geometry Standard configurations Degrees of freedom Grippers and graspability conditions Goal.
CSCE 641: Forward kinematics and inverse kinematics Jinxiang Chai.
M. Farber and D. Schütz | Homology of Planar Polygon Spaces | Presented by Yair Carmon Based on the paper by M. Farber and D. Schütz Presented by Yair.
Rational Trigonometry Applied to Robotics
CSCE 641: Forward kinematics and inverse kinematics Jinxiang Chai.
Ch. 4: Velocity Kinematics
BINARY MORPHOLOGY and APPLICATIONS IN ROBOTICS. Applications of Minkowski Sum 1.Minkowski addition plays a central role in mathematical morphology 2.It.
CSCE 689: Forward Kinematics and Inverse Kinematics
Chapter 5: Path Planning Hadi Moradi. Motivation Need to choose a path for the end effector that avoids collisions and singularities Collisions are easy.
Inverse Kinematics Jacobian Matrix Trajectory Planning
Introduction to ROBOTICS
KINEMATICS ANALYSIS OF ROBOTS (Part 1) ENG4406 ROBOTICS AND MACHINE VISION PART 2 LECTURE 8.
Advanced Graphics (and Animation) Spring 2002
Constraints-based Motion Planning for an Automatic, Flexible Laser Scanning Robotized Platform Th. Borangiu, A. Dogar, A. Dumitrache University Politehnica.
Key Frame Animation and Inverse Kinematics
KINEMATICS ANALYSIS OF ROBOTS (Part 2)
1 C03 – Advanced Robotics for Autonomous Manipulation Department of Mechanical EngineeringME 696 – Advanced Topics in Mechanical Engineering.
CS B659: Principles of Intelligent Robot Motion Configuration Space.
The City College of New York 1 Dr. Jizhong Xiao Department of Electrical Engineering City College of New York Inverse Kinematics Jacobian.
CSCE 441: Computer Graphics Forward/Inverse kinematics Jinxiang Chai.
Review: Differential Kinematics
Chapter 7: Trajectory Generation Faculty of Engineering - Mechanical Engineering Department ROBOTICS Outline: 1.
M. Zareinejad 1. 2 Grounded interfaces Very similar to robots Need Kinematics –––––– Determine endpoint position Calculate velocities Calculate force-torque.
Kinematics. The function of a robot is to manipulate objects in its workspace. To manipulate objects means to cause them to move in a desired way (as.
Just a quick reminder with another example
ECE 450 Introduction to Robotics Section: Instructor: Linda A. Gee 10/07/99 Lecture 11.
Outline: Introduction Solvability Manipulator subspace when n<6
Trajectory Generation
Chapter 3 Differential Motions and Velocities
Robotics Chapter 3 – Forward Kinematics
Velocity Propagation Between Robot Links 3/4 Instructor: Jacob Rosen Advanced Robotic - MAE 263D - Department of Mechanical & Aerospace Engineering - UCLA.
Kinematics 제어시스템 이론 및 실습 조현우
FREE TOPOLOGICAL GROUPS AND FUNCTORIAL FIBREWISE CONSTRUCTIONS
MT411 Robotic Engineering
CSCE 441: Computer Graphics Forward/Inverse kinematics
Trajectory Generation
Spatcial Description & Transformation
INVERSE MANIPULATOR KINEMATICS
LOWER ESTIMATES FOR THE TOPOLOGICAL COMPLEXITY
Continuum Mechanics (MTH487)
Basilio Bona DAUIN – Politecnico di Torino
Direct Manipulator Kinematics
Automatically trivial fibrations
Zaid H. Rashid Supervisor Dr. Hassan M. Alwan
ROTATIONS & TRANSLATIONS
CHAPTER 2 FORWARD KINEMATIC 1.
Introduction to manipulator kinematics
Mobile Robot Kinematics
Motion in Two Dimensions
Distinct Distances in the Plane
ON THE LUSTERNIK-SCHNIRELMANN CATEGORY APPROACH
Matrix Methods in Kinematics
CSCE 441: Computer Graphics Forward/Inverse kinematics
Inverse Kinematics 12/30/2018.
Robot Kinematics We know that a set of “joint angles” can be used to locate and orientate the hand in 3-D space We know that the joint angles can be combined.
FREE TOPOLOGICAL GROUPS AND FUNCTORIAL FIBREWISE CONSTRUCTIONS
Motion in Two Dimensions
How to Get Where you Want to Go
Maths for Signals and Systems Linear Algebra in Engineering Lectures 13 – 14, Tuesday 8th November 2016 DR TANIA STATHAKI READER (ASSOCIATE PROFFESOR)
Outline: 5.1 INTRODUCTION
Petar Pavešić, University of Ljubljana
TOPOLOGICAL COMPLEXITY OF FIBRATIONS AND COVERING MAPS
Chapter 4 . Trajectory planning and Inverse kinematics
Chapter 3. Kinematic analysis
Robotics 1 Copyright Martin P. Aalund, Ph.D.
CSE 599 – Arm Planning Joseph Xu.
Presentation transcript:

TOPOLOGICAL COMPLEXITY OF KINEMATIC MAPS ABSTRACT TOPOLOGICAL COMPLEXITY OF KINEMATIC MAPS Petar Pavešić, University of Ljubljana In a mechanical device, like a robot arm, the forward kinematic map relates the configuration space of the joints to the working space of the device. A typical kinematic map is smooth but with certain singularities, and it admits only partial inverses. We are going to describe a general setting for the study of the motion planning problem in this context, and discuss the complexity of some simple joint configurations. Lisboa - 2015 1

TOPOLOGICAL COMPLEXITY TOPOLOGICAL COMPLEXITY OF KINEMATIC MAPS INTRODUCTION TOPOLOGICAL COMPLEXITY Motion planning in configuration spaces: 𝒞 - configuration space of a mechanical device Find robust rules that product paths connecting any two given configurations in 𝒞. 𝑐 𝑖𝑛𝑖𝑡 , 𝑐 𝑓𝑖𝑛 ∈𝒞×𝒞 query 𝜌:𝒞×𝒞→𝑃𝒞, 𝜌 𝑐 𝑖𝑛𝑖𝑡 , 𝑐 𝑓𝑖𝑛 0 = 𝑐 𝑖𝑛𝑖𝑡 , 𝜌 𝑐 𝑖𝑛𝑖𝑡 , 𝑐 𝑓𝑖𝑛 1 = 𝑐 𝑓𝑖𝑛 roadmap (Farber) Global roadmap exists ⇔ 𝒞 is contractible. Consider partial roadmaps, defined on subsets of 𝒞×𝒞. 𝑻𝑪 𝓒 := minimal number of partial roadmaps, needed to satisfy all queries. 2

Partial roadmaps are partial sections of 𝜋. TOPOLOGICAL COMPLEXITY OF KINEMATIC MAPS INTRODUCTION 𝒞×𝒞 𝑃𝒞 𝜋 𝛼 (𝛼 0 ,𝛼 1 ) 𝜌 Partial roadmaps are partial sections of 𝜋. 𝒬 𝜋 is a fibration 𝑇𝐶 𝒞 is the Schwarz genus of 𝜋 In particular, 𝑇𝐶 𝒞 is a homotopy invariant of 𝒞. Extensive computations of 𝑇𝐶 𝒞 based on explicit roadmaps (upper bounds) and cohomology obstructions (lower bounds). 3

Actual motion planning is more complex: TOPOLOGICAL COMPLEXITY OF KINEMATIC MAPS INTRODUCTION Actual motion planning is more complex: configuration space 𝒞 (position of joints) robot device 𝑓 – forward kinematic map working space 𝒲 (spatial position of the device) universal joint planar joints  (x,y,z)    𝒞= 𝑇 2 𝒞= 𝑇 2 = ( 𝑒 𝑖𝜑 , 𝑒 𝑖𝜓 )∈ 𝑆 1 × 𝑆 1 𝒲= 𝑆 2 𝒲= (𝑥,𝑦)∈ ℝ 2 𝑟 2 ≤ 𝑥 2 + 𝑦 2 ≤ 𝑅 2 𝑥=𝑟 cos 𝜓 cos 𝜑 𝑦=𝑟 cos 𝜓 sin 𝜑 𝑧=𝑟 sin 𝜓 𝑓 𝑒 𝑖𝜑 , 𝑒 𝑖𝜓 =(𝑅 cos 𝜑+𝑟 cos 𝜑+𝜓 , 𝑅 sin 𝜑+𝑟 sin 𝜑+𝜓 ) 4

𝒞= 𝑇 3 robot arm wrist 𝒲=𝑆𝑂(3) TOPOLOGICAL COMPLEXITY OF KINEMATIC MAPS INTRODUCTION robot arm wrist 𝒞= 𝑇 3 𝒲=𝑆𝑂(3) Forward kinematic map 𝑓:𝑇 3 ⟶𝑆𝑂 3 is given In terms of Euler angles as 𝑓 𝑒 𝑖𝜑 , 𝑒 𝑖𝜓 , 𝑒 𝑖𝜃 = cos 𝜑 cos 𝜓 cos 𝜃 − sin 𝜑 sin 𝜃 − cos 𝜃 sin 𝜑 − cos 𝜑 cos 𝜓 sin 𝜃 cos 𝜑 sin 𝜃 + cos 𝜓 cos 𝜃 sin 𝜑 cos 𝜑 cos 𝜃 − cos 𝜓 sin 𝜑 sin 𝜃 cos 𝜑 sin 𝜓 sin 𝜑 sin 𝜓 − cos 𝜓 sin 𝜓 sin 𝜓 sin 𝜃 cos 𝜓 5

(Dranishnikov) Global roadmap exists ⇒ 𝑓 admits a section. TOPOLOGICAL COMPLEXITY OF KINEMATIC MAPS INTRODUCTION MOTION PLANNING 𝑓:𝒞⟶𝒲 forward kinematic map from a configuration space 𝒞 to a working space 𝒲 Find robust rules to move the device from a given joint configuration to a desired position in the working space. 𝑐 𝑖𝑛𝑖𝑡 , 𝑤 𝑓𝑖𝑛 ∈𝒞×𝒲 query 𝜌:𝒞×𝒲→𝑃𝒞, 𝜌 𝑐 𝑖𝑛𝑖𝑡 , 𝑤 𝑓𝑖𝑛 0 = 𝑐 𝑖𝑛𝑖𝑡 , 𝑓(𝜌 𝑐 𝑖𝑛𝑖𝑡 , 𝑤 𝑓𝑖𝑛 1 )= 𝑤 𝑓𝑖𝑛 roadmap (Dranishnikov) Global roadmap exists ⇒ 𝑓 admits a section. Proof: given 𝜌:𝒞×𝒲→𝑃𝒞, 𝑤⟼𝜌( 𝑐 0 ,𝑤)(1) is a section of 𝑓. If 𝑓 does not admit a section consider partial roadmaps, defined on subsets of 𝒞×𝒲. 𝑻𝑪 𝒇 := minimal number of partial roadmaps, needed to satisfy all queries. 6

partial sections of (1×𝑓)𝜋. TOPOLOGICAL COMPLEXITY OF KINEMATIC MAPS INTRODUCTION Assume 𝒞 ,𝒲,𝒬 are ENRs: 𝒞×𝒲 𝑃𝒞 (1×𝑓)𝜋 𝛼 (𝛼 0 ,𝑓(𝛼 1 )) 𝜌 Partial roadmaps are partial sections of (1×𝑓)𝜋. 𝒬 If 𝑓 is a fibration, then 𝑇𝐶 𝑓 is the Schwarz genus of(1×𝑓)𝜋. However, forward kinematic maps that arise in practice: (1) do not admit inverse kinematic maps (i.e. sections), and (2) are smooth but always have singular points (where Jacobian does not have maximal rank). Therefore are not fibrations. (D. Gottlieb, 1986, 1988) In general 𝑇𝐶 𝑓 ≥ Schwarz genus of(1×𝑓)𝜋 (and the difference can be arbitrarily big). Problem: compute 𝑇𝐶 𝑓 for basic robot arm configurations. 7

(1) Assume 𝑓:𝒞⟶𝒲 admits section 𝑠:𝒲⟶𝒞: TOPOLOGICAL COMPLEXITY OF KINEMATIC MAPS SPECIAL CASES Observe: 𝑇𝐶 1 𝑋 =𝑇𝐶 𝑋 . Two special cases: (1) Assume 𝑓:𝒞⟶𝒲 admits section 𝑠:𝒲⟶𝒞: Then, to move from 𝒸∈𝒞 to 𝑤∈𝒲 it is sufficient to move from 𝒸 to 𝑠 𝑤 in 𝒞, and to move from 𝑤 to 𝑤′ in 𝒲 it is sufficient to more from 𝑠(𝑤) to 𝑤′. 𝑓 sectioned ⟹ 𝑇𝐶 𝒲 ≤𝑇𝐶 𝑓 ≤𝑇𝐶 𝒞 (2) Assume 𝑓:𝒞⟶𝒲 fibration: Then, to move from 𝒸∈𝒞 to 𝑤∈𝒲 it is sufficient to find a path in 𝒲 from 𝑓(𝒸) to 𝑤, and lift it to 𝒞 starting at 𝒸. 𝑓 fibration ⟹ 𝑇𝐶 𝑓 ≤𝑇𝐶 𝒲 8

(3) 𝑇𝐶 𝑓 𝒬∪𝒬′ ≤𝑇𝐶 𝑓 𝒬 +𝑇𝐶 𝑓 𝒬′ TOPOLOGICAL COMPLEXITY OF KINEMATIC MAPS RELATIVE COMPLEXITY In general, avoid singularities and split into subsets that admit local sections. RELATIVE COMPLEXITY 𝑓:𝒞⟶𝒲, 𝒬⊆𝒞×𝒲 𝑻𝑪 𝒇 𝓠 = minimal number of partial roadmaps, needed to satisfy all queries in 𝒬. Properties: (1) 𝑇𝐶 𝑓 Γ(𝑓) =1 (2) 𝒬⊆ 𝒬 ′ ⟹ 𝑇𝐶 𝑓 𝒬 ≤𝑇𝐶 𝑓 𝒬′ (3) 𝑇𝐶 𝑓 𝒬∪𝒬′ ≤𝑇𝐶 𝑓 𝒬 +𝑇𝐶 𝑓 𝒬′ (4) If 𝒬 ′ can be horizontally deformed into 𝒬, then 𝑇𝐶 𝑓 𝒬 ≥𝑇𝐶 𝑓 𝒬′ In particular, if 𝒬 ′ can be horizontally deformed to a subset 𝒬, then 𝑇𝐶 𝑓 𝒬 =𝑇𝐶 𝑓 𝒬′ . 9

𝒲 𝑟 ≔ regular values of 𝑓, 𝒞 𝑟 := 𝑓 −1 ( 𝒲 𝑟 ) TOPOLOGICAL COMPLEXITY OF KINEMATIC MAPS RELATIVE COMPLEXITY 𝑓:𝒞⟶𝒲 smooth 𝒲 𝑟 ≔ regular values of 𝑓, 𝒞 𝑟 := 𝑓 −1 ( 𝒲 𝑟 ) 𝑇𝐶 𝑓 𝒞 𝑟 × 𝒲 𝑟 ≤𝑇𝐶 𝑓 𝒲 𝑟 𝑇𝐶 𝑓 𝒞× 𝒲 𝑟 ≤𝑐𝑎𝑡(𝒞× 𝒲 𝑟 ) 𝑓:𝒞⟶𝒲 admits a partial section 𝑠 over 𝒲 ′ , 𝒞 ′ := 𝑓 −1 ( 𝒲 ′ ) 𝑇𝐶 𝒲 𝒲× 𝒲 ′ ≤ 𝑇𝐶 𝑓 𝒞× 𝒲 ′ ≤ 𝑇𝐶 𝒞 𝒞× 𝑠(𝒲 ′ ) Moreover, if 𝒞 ′ can be deformed to 𝑠(𝒲 ′ ), then 𝑇𝐶 𝑓 𝒞× 𝒲 ′ =𝑇𝐶 𝒞 𝒞× 𝑠(𝒲 ′ ) and if 𝒲 ′ ⊆ 𝒲 𝑟 , then 𝑇𝐶 𝑓 𝒞′× 𝒲 ′ =𝑇𝐶( 𝒲 ′ ) If 𝒞 is a topological group, then 𝑇𝐶 𝒞 𝒞× 𝑠(𝒲 ′ ) =𝑇𝐶 𝒞 =𝑐𝑎𝑡 𝒞 . 10

Kinematic map is sectioned and is regular except for 𝜓=0,𝜋. TOPOLOGICAL COMPLEXITY OF KINEMATIC MAPS COMPUTATIONS planar joints Kinematic map is sectioned and is regular except for 𝜓=0,𝜋.   ⟹ 2=𝑇𝐶 𝑆 1 ×𝐼 ≤𝑇𝐶 𝑓 ≤𝑇𝐶 𝑇 2 =3 Furthermore, for 𝒞 ′ =𝑆 1 ×0 we have 𝑇𝐶 𝑓 ≥𝑇𝐶 𝑓 𝑇 2 ×𝑓( 𝒞 ′ )=𝑇𝐶 𝑇 2 𝑇 2 × 𝒞 ′ =3. Therefore 𝑇𝐶 𝑓 =3. universal (Cardan) joint Kinematic map is regular except for 𝜓=± 𝜋 2 , and is sectioned over 𝑆 2 − poles .   (x,y,z) Previous results yield 3≤𝑇𝐶 𝑓 ≤5,… … which may be improved to 4≤𝑇𝐶 𝑓 ≤5, by a careful analysis of extensions of roadmaps around singular points. 11

These are matrices of the form TOPOLOGICAL COMPLEXITY OF KINEMATIC MAPS COMPUTATIONS robot arm wrist Singular values of 𝑓 are rotations whose Z-axis coincides with the Z-axis of the reference frame. These are matrices of the form corresponding to points on a circle in 𝑆𝑂(3). cos 𝜃 − sin 𝜃 sin 𝜃 cos 𝜃 0 0 0 0 1 Singular points are of the form 𝑒 𝑖𝜑 ,1, 𝑒 𝑖𝜃 and 𝑒 𝑖𝜑 ,−1, 𝑒 𝑖𝜃 , corresponding to points in a disjoint union of two two-dimensional tori in 𝑇 3 . 𝑓 admits a section on the complement of singular points, and the previous estimates yield 4≤𝑇𝐶 𝑓 ≤6. 12

TOPOLOGICAL COMPLEXITY OF KINEMATIC MAPS Further results We implicitly used cohomological lower bounds for the topological complexity of various configuration and working spaces. One can also consider the diagram for partial liftings and use the standard approach to derive the estimate 𝒬 𝒞×𝒲 𝑃𝒞 (1×𝑓)𝜋 𝜌 𝑇𝐶 𝑓 ≥ nil ( Ker (1,𝑓) ∗ : 𝐻 ∗ (𝒞×𝒲)⟶ 𝐻 ∗ (𝒞) ) (Also follows from the fact that 𝑇𝐶 𝑓 is bigger than the Schwarz genus of (1×𝑓)𝜋.) 13

THANKS! 14