Академический Документы
Профессиональный Документы
Культура Документы
Abstract
Humanoid Robot Kinematics modeling for robotics real-time
applications requires computationally efficient solutions of the inverse kinematics
problem. As the complexity increases, the need for more elegant formulations of
the equations of motion becomes a paramount issue. This paper investigates
analytical methods, using techniques and notation from the theory of Lie groups
and Lie algebras, to develop geometric algorithms for the kinematical analysis,
modeling and simulation of humanoid robots.
We introduce a new closed-form analytical solution, geometrically
meaningful and numerically stable, based on the PoE (Product of Exponentials
formula), which allows us to develop a geometric algorithm to solve the
humanoid inverse kinematics problem, built upon the Paden-Kahan subproblems.
Our approach to path planning the WBT (Whole Body Trajectory) for
humanoid robots computes collision-free trajectories by the development of the
new FM3 (Fast Marching Method Modified) algorithm. We present a general and
analytical solution for the inverse kinematics, satisfying the balance constraints of
the ZMP (Zero Moment Point) trajectory, based on a Humanoid SKD (Sagital
Kinematics Division) approach; this is the humanoid OSG (One Step to Goal)
algorithm. The motion of the humanoid consists on applying successive goals
from the WBT, to the inverse kinematics algorithm.
Our results are illustrated in studies with the 21 DoF (degree-offreedom) humanoid robot RH0 (Universidad Carlos III de Madrid). The algorithm
is presented along with computed simulation examples.
General Index
ABSTRACT ............................................................................................. 2
GENERAL INDEX ................................................................................... 3
1. INTRODUCTION ................................................................................. 4
2. BACKGROUND ................................................................................... 5
2.1 Geometric Methods in Robotics - Lie Groups & Algebras Screw theory. 5
2.2 The PoE (Product of Exponentials) formula Forward Kinematics........... 6
2.3 The PoE Inverse Kinematics by Paden-Kahan subproblems. ....................... 7
2.4 FMM (Fast Marching Methods).......................................................................... 9
6 CONCLUSION .................................................................................... 17
REFERENCES ....................................................................................... 18
1. Introduction
The Humanoid robot kinematics modeling presents a formidable
computational challenge due to the high number of degrees of freedom, complex
kinematics models, and balance constraints ([10],[12],[20]). Traditionally, inverse
kinematics algorithms have been developed through a numerical approach based
on a variety of optimization and Heuristic methods ([7],[11],[15]). Conversely,
this paper presents a new analytical closed-form method for the solution of the
humanoid robot kinematics modeling using techniques from the theory of Lie
Groups, such us the PoE (Product of exponentials formula), allowing elegant, fast
and efficient calculations, even for real-time applications.
One objective is to path planning the collision-free WBT (Whole
Body Trajectory) for the humanoid, by developing the FM3 (Fast Marching
Method Modified) algorithm, based on Level Set and Fast Marching Methods
([1]), which are techniques for analyzing evolving interfaces in computational
geometry.
Another objective of this paper is to introduce the OSG (One Step to
Goal) algorithm, for generating statically-stable motions through the solution of
the humanoid inverse kinematics, by imposing balance constraints of the ZMP
(Zero Moment Point) trajectory, upon incremental goals obtained during the WBT
search. The OSG algorithms is developed from a divide and conquer approach
called SKD (Sagital Kinematics Division), which from the kinematical point of
view, considers the humanoid to be divided by the sagital plane into two
synchronized manipulators. On so doing, the inverse kinematics problem is solved
in an analytical and complete (for all DoF) way, using techniques from the Lie
groups, these are, PoE and the Paden-Kahan subproblems ([2],[21],[22]).
The development of general-purpose motion generation tools and
techniques has been based on the use of virtual humanoid robot platforms ([10],
[12]). In a similar way, for testing our results with the 21 DoF humanoid robot
RH0 (Universidad Carlos III de Madrid), we use a graphical simulation
environment with the Humanoid based on the H-Anim (Humanoid Animation)
definition, and VRML (Virtual Reality Modeling Language) for the building of
the virtual world.
2. Background
We provide a description of the rigid body motions using the tools of
Lie algebras and Screw theories for the study of robot kinematics ([2], [3], [5],
[16]). From these theories, we present the tools we use for the analytical solution
of the humanoid inverse kinematics problem, these are: the POE (product of
exponentials) and the Paden-Kahan subproblems ([21],[22]).
The question of motion planning and obstacle avoidance for the
humanoid is analyzed through the use of Fast Marching Methods ([1]).
R p
g=
R = Orientatio n ; p = position
0
1
^
3
^=
, are the TWIST Coordinate s of ^
0
^
1
, 2^ = 1^ 2^ 2^ 1^
5
________
1
= 2 ^ =
3
3
2
3
0
2
1 skew symmetric matrix so (3)
0
R
g=
0
p e
=
1 0
^
( I e ) ( ) + T
0
1
________
the axis. For a prismatic joint, is a unit vector pointing in the direction of
translation. Then, the product of exponentials formula for the manipulator forward
kinematics is gst().
q
i = = i i ; twist for revolute
i
i
i = = ; twist for prismatic
0
g st ( ) = e
1 ^ 1
2 ^ 2
Le
n ^ n
g st (0 )
________
c = r +
(
T
(
2
)
T
1
2
2
T
1
1 = 1
1
v
2
u
2
)
T
2
2
T
1
2
2
Once we get c for the second subproblem, we can apply the first
Paden-Kahan subproblem to get the solutions for 1 and 2. Beware that there
might exist two solutions for c , each one of them gives a solution for 1 and 2.
= 0 cos
2 u' v'
'2 = 2 T ( p q)
u' = u Tu
v' = v T v
=
From the solution for the third subproblem is clear that it could lead to
multiple solutions as well.
Once these subproblems are solved, the full inverse kinematics
problem can be solved by reducing it into appropriate subproblems ([2]), which
are geometrically meaningful and numerically stable.
________
dx = F ( dT )
1 = F dT
dx
T F = 1
Figura 5: One dimension Eikonal Equation example.
max( D T , D T ,0 )
1
=
y
+y
2
Fij
+ max( Dij T , Dij T ,0 )
T ( + h , t ) T ( , t )
Dij+ T
h
T ( , t ) T ( h , t )
Dij T
h
x
ij
+x
ij
________
10
3. Humanoid Model
RH0 Robot
The Universidad Carlos III de Madrid has lunched RH0 humanoid
robotics project. The mechanical design concepts of RH0 are light, compact, but
performable for application tasks like working cooperation with humans.
Several architecture humanoid robotics platforms have been
developed in recent years ([10], [12]). There are many advantages for an open
architecture software and hardware to obtain the consistency between the
simulator and a real humanoid robot, and therefore we work on the same idea,
even though we did not joint to any specific architecture. The models of robots
and their working environment are described by VRML97 format that is extended
for a humanoid animation by h-anim working group [19].
________
11
T F = 1
The new version of FMM is the recently developed algorithm FM3,
which stands for Fast Marching Method Modified. This new algorithm overcomes
the problem of having slow numeric methods trough the building of a lineal
scheme for approximating the Eikonal equation, directly resoluble, and therefore
________
12
computationally very fast and suitable for real time applications. This method
allows to always find out a cuasi-optimal humanoid WBT path, whatever the
nature of the obstacles, this is, the FM3 solves the humanoid collision-free path
inside a world with regular, irregular and even non-convex geometry obstacles.
The FM3 solves the Eikonal equation to produce the Arrival Function
T. Then, given the goal point, explicit construction of the shortest path comes
through back propagation following the maximum negative gradient. An efficient
scheme for R3 is presented hereafter, being h the integration step ant t the time.
1
Fij
T ( h, y ) T ( , y )
Dij T Abs
T ( x, y , z ) = min
( x, y,z )
F ( x , y , z ) d
________
13
5. Humanoid
Kinematics Motion
We introduce the OSG (One Step to Goal) algorithm, for generating
statically-stable humanoid motions based on incremental goals obtained from the
WBT path planning, by imposing balance constraints of the ZMP (Zero Moment
Point) trajectory, upon the inverse kinematics algorithm. The OSG algorithm is
developed from the previously presented Humanoid model SKD (Sagital
Kinematics Division) humanoid model. The humanoid inverse kinematics motion
problem is solved in an analytical and complete way, using techniques from the
Lie groups (PoE and the Paden-Kahan subproblems) ([2]).
________
14
Figure 3: The five phases for the OSG (One Step to Goal)
Algorithm: 1-Orientate, 2-Tilt, 3-Elevate, 4-Lean and 5-Balance.
________
15
g st ( ) = e1 1 e 2 2 L e12 12 g st (0 )
^
q
i = = i i
i
________
16
9^9
= e
pq
K 3
P
9
7^7
q' = e
8^8
K 2
p' P
7 ,8
1^1
10^10
10^10
q' = e
11^11
12^12
K 2
e11 11 p' P
10,11
K 1
That is it!
________
17
6 Conclusion
We believe that abstraction saves time in the end, in return for an
initial investment in learning some mathematics. Therefore, this paper presents a
slightly more abstract formulation, for humanoid robot kinematics modeling using
the theory of Lie groups and Lie algebras. These techniques allow elegant, fast
and efficient calculations for real-time applications.
The main development of this work is to introduce the new OSG (One
Step to Goal) algorithm, for generating statically stable humanoid motions
through the solution of the inverse kinematics problem, by imposing balance
constraints of the ZMP trajectory. We present the new approach called SKD
(Sagital Kinematics Division), which from the kinematical point of view,
considers the humanoid divided by its sagital plane into two synchronized
manipulators. We solve the problem by an analytical method using techniques
from the theory of Lie Groups, such us the PoE (Product of exponentials
formula), and the Paden-Kahan subproblems.
Besides, this paper investigates the path planning problem for the
collision-free WBT (Whole Body Trajectory) of the humanoid, by developing the
new FM3 (Fast Marching Method Modified) algorithm, based on techniques for
analyzing evolving interfaces in computational geometry.
For testing our results we worked with the 21 DoF humanoid robot
RH0 (Universidad Carlos III de Madrid), we use a graphical simulation
environment based on the H-Anim (Humanoid Animation) definition, and VRML
for the building of the virtual world.
We hope that with the use of new mathematical developments, goallevel planning algorithms and interactive simulation software, the capabilities of
humanoid robotics technology will enter mainstream society during the next
several decades.
References
[1] J.A. Sethian. Level Set Mehods and Fast Marching Methods,
Evolving Interfaces in ComputationalCambridge University
Press. 1999.
[2] R.M. Murray, Z. Li, and S.S. Sastry, A Mathematical Introduction to
Robot Manipulation. Boca Raton, FL: CRC Press, 1993.
[3] R. Featherstone, Robot Dynamics Algorithms. Boston: Kluwer,
1987.
[4] J.C. Latombe. Robot Motion Planning. Kluwer Academic
Publishers, Boston, USA, 1991.
[5] F.C. Park, J.E. Bobrow, and S.R. Ploen, \A Lie group formulation
of robot dynamics," Int. J. Robotics Research. Vol. 14, No. 6, pp.
609-618, 1995.
[6] Ronald C. Arkin. Behavior Based Robotics. The MIT Press 1998.
[7] Liegeois, A., Automatic supervisory control of the configuration and
behavior of multibody mechanisms. IEEE Transactions on Systems,
Man, and Cybernetics, 1977. 7(12): p. 868-871.
[8] G. Rodriguez, A. Jain, and K. Kreutz-Delgado, \Spatial operator
algebra for multibody system dynamics," J. Astronautical Sciences.
Vol. 40, No. 1, pp. 27-50, 1992.
[9] M. Vukobratovic and D. Juricic, Contribution to the Synthesis of
Biped Gait, IEEE Tran. On Bio-Medical Engineering, Vol. 16, No.
1, pp. 1-6, 1969.
[10] F. KANEHIRO, K. FUJIWARA, et al. Open Architecture
Humanoid Robotics Platform. Proceedings of the 2002 IEEE
International Conference on Robotics & Automation Washington,
DC May 2002.
18
Referencias
19
[11] Baillieul, J. Kinematic programming alternatives for redundant
manipulators. in IEEE International Conference on Robotics and
Automation. 1985.
[12] J.J. Kuffner, K. Nishiwaki, S. Kagami, M. Inaba, and H. Inoue.
Motion Planning for Humanoid Robots. In Proc. 11th Intl Symp.
of Robotics Research (ISRR 2003).
[13] Matlab 6.5, The MathWorks, Inc., 2003.
[14] M. Hardt, K. Kreutz-Delgado, and J. William Helton, \Minimal
Energy Control of a Biped Robot with Numerical Methods and a
Recursive Symbolic Dynamic Model," Proc. 37th IEEE Conference
on Decision and Control, pp. 413{6, 1998.
[15] Klein, C.A., C. Chu-Jenq, and S. Ahmed, A new formulation of the
Extended Jacobian method and its use in mapping algorithmic
singularities for kinematically redundant manipulators. IEEE
Transactions on Robotics and Automation, 1995. 11(1): p. 50-55.
[16] R.S. Ball. The Theory of Screws. Cambridge University Press,
Cambridge, 1900.
[17] R. Brockett. Robotic manipulators and the product of exponential
formula. In P. Fuhrman ed. Proc. Mathematical Theory of
Networks and Systems pp. 120129, 1984.
[18] F. Hausdorff. Die Symbolische exponential formel in den gruppen
theorie. Berichte de Sachicen Akademie de Wissenschaften (Math
Phys Klasse) vol. 58, pp. 1948, 1906.
[19] http://www.h-anim.org/. HUMANOID ANIMATION
WORKING GROUP.
[20] Inoue H., Tachi S., Tnie K., Yokoi K., Huirai S., Hirukawa H., Hirai
K., Nakayama S., Sawada K., Nishiyama T., Miki O., Itoko T., Inaba
H., Sudo M., University of Tokyo, MEL, ETL, Honda R&D Co.
Ltd., Matsuita Ltd, Kawasaki Ltd., Fanuc Ltd., HRP: Humanoid
Robotics Project of MITI, Proc. First IEEE-RAS International
Conference on Humanoid Robots (HUMANOID2000), Sep. 2000.
[21] B. Paden. Kinematics and Control Robot Manipulators. PhD thesis,
Department of Electrical Engineering and Computer Sciences,
University of California, Berkeley, 1986.
[22] W. Kahan. Lectures on computational aspects of geometry.
Department of Electrical Engineering and Computer Sciences,
University of California, Berkeley, 1983.