Intelligent Robotics Lecture
Tomáš Pajdla 2008
1
Advanced Robotics
Lecture 1
2
ROBOT R.U.R. (Rossum's Universal Robots) by Karel Čapek. Rossum's robots are biological creations that have skin mixed in a vat, and their nerves and digestive tracts spun on spindles, and are then assembled like automobiles. They resemble more modern conceptions of manmade life forms such as the Replicants in Blade Runner.
3
4
ROBOT
=
A
GENERAL
MANIPULATOR
5
We shall learn how to solve dvanced kinematic problems for manipulators with 6-degrees-of-freedom.
The general solution to this problem exists. 1. Kinematic calibration 2. Motion planning 3. Eye-hand systems
6
Industrial Robotic Applications
7
8
9
10
Precision for industry Low (e.g. manipulation) ± 5 mm in the whole working space ± 0.5 mm locally … often available
High (e.g. laser welding) ± 0.5 mm in the whole working space ± 0.05 mm locally … often not available
11
Modeling kinematics – calibration – absolute acuracy ± 0.05 mm
Robot-Vision calibration (courtesy Neovision s.r.o.)
13
14
Precision for robotic surgery
15
http://www.cts.usc.edu/rsi-article-robotputsuscatforefront.html
16
17
Two kinds of manipulators
1. Serial manipulators 2. Parallel manipulators
18
Serial manipulators
J
KUKA manipulator
19
Serial manipulators
Stäubli (courtesy Neovision s.r.o.)
Mitsubishi (courtesy Neovision s.r.o.)
1. Direct kinematic task – easy 2. Inverse kinematic task – difficult
20
Parallel manipulators
J
Stewart-Gough Platform
21
Parallel manipulators
Sliding Star (courtesy of Prof. Valášek, CTU Prague)
1. Direct kinematic task – difficult 2. Inverse kinematic task – easy
22
Kinematics in robotics
Three main problems
1. Direct kinematic task (přímá kinematická úloha) 2. Inverse kinematic task (inverzní kinematická úloha) 3. Kinematic calibration (kalibrace kinematiky)
23
Direct kinematic task
flange frame
z
x
world frame
y
24
Inverse kinematic task
flange frame
z
x
world frame
y
25
Kinematic calibration
z
x
world frame
y
26
Kinematic Calibration
27
Robot Calibration
28
Kinematic Calibration
29
30
32
Solving kinematic tasks
33
Solving kinematic tasks 1968 Donald L. Pieper (Ph.D. thesis) The inverse kinematics of any serial manipulator with six revolute joints, and with three consecutive joints intersecting, can be solved in closedform, i.e., analytically.
1989 M. Raghavan, B. Roth. Kinematic Analysis of the 6R Manipulator of General Geometry. Int. Symp. Robotics. Research. Pp. 314-320, Tokyo 1989/1990. A general technique for computing inverse kinematics for any serial manipulator with six revolute joints. … leads to solving an algebraic equation of degree 16.
34
Solving kinematic tasks Algebraic equation of degree 16 … up to 16 solutions
4 typical solutions
35
Solving kinematic tasks
36
Stäubli TX-90 – Geometry
37
Kinematic model
flange frame z x z y
x
world frame
y
38
39
The Standard Kinematic model in Denavit-Hartenberg Convention Stäubli TX 90
TX-90 (6 axis, RRRRRR) [Staubli] α -1.5708 0.0 -1.5708 1.5708 -1.5708 0.0 a 50.0 425.0 0.0 0.0 0.0 0.0 θ 0.0 0.0 0.0 0.0 0.0 0.0 d 350.0 50.0 0.0 425.0 0.0 100.0
6 non-trivial parameteres
40
The Standard Kinematic model in Denavit-Hartenberg Convention ABB IBR 140
41
The Standard Kinematic model in Denavit-Hartenberg Convention ABB IBR 140
IBR-140 (6 axis) [ABB] α -1.5708 0.0 -1.5708 1.5708 -1.5708 0.0 a 70.0 360.0 0.0 0.0 0.0 0.0 θ 0.0 0.0 0.0 0.0 0.0 0.0 d 352.0 0.0 0.0 380.0 0.0 65.0
5 non-trivial parameteres
42
The Standard Kinematic model in Denavit-Hartenberg Convention Stäubli TX 90
RV-6S (6 axis, RRRRRR) [Mitsubishi] α -1.5708 0.0 -1.5708 1.5708 -1.5708 0.0 a 85.0 280.0 100.0 0.0 0.0 0.0 θ 0.0 0.0 0.0 0.0 0.0 0.0 d 350.0 0.0 0.0 315.0 0.0 85.0
6 non-trivial parameteres
43
Special versus General Mechanisms Special simple & tractable
α -1.5708 0.0 -1.5708 1.5708 -1.5708 0.0 a 70.0 360.0 0.0 0.0 0.0 0.0 θ d 352.0 0.0 0.0 380.0 0.0 65.0 α -1.42 0.10 -1.57 1.58 -1.59 0.07
×
General complicated & hard
a 70.1 360.0 0.2 0.1 0.4 0.2 θ - (+0.2) - (+0.1) - (- 0.3) - (+0.1) - (- 0.1) - (- 0.2) d 352.0 0.2 0.3 380.2 0.1 65.1
6 non-trivial parameters
×
18 (+6) non-trivial parameters
High precision → Small misalignments important → General mechanisms
44
45
46
47
48
49
50
51
SOLVING 1 ALGEBRAIC EQUATION 1 equation, 1 variable → companion matrix → eigenvalues
... a simple rule
It works when eig works, i.e. order 100 in Matlab is often OK.
52