Documente Academic
Documente Profesional
Documente Cultură
1 23
1 23
1 Introduction
Wheeled Mobile Robots (WMRs) are most often
assumed to be capable of rolling without slipping, and
modeled as nonholonomic dynamic systems [8, 9, 28,
36, 39]. The guidance and the control of the WMR
is influenced by the features of terrain on which the
robot operates. Due to various effects such as slipping, scrubbing, sliding, deformability or flexibility of
the wheels, the ideal constraints are never strictly satisfied. For autonomous mobile robots driving across
soft soils, like sand, loose dirt or snow it is essential to deal with the dynamic effects occurring at the
wheelterrain contiguity. Wong [37] gives a fundamental understanding of the critical factors affecting
2 Dynamics Equations
Let us focus on a class of WMRs, whose phase
constraints of motion (e.g. pure rolling, non-slipping
conditions at the contact point of each wheel with the
ground) are fully transgressed. Let q Rn denote
generalized coordinates of the mobile robot. The Pfaffian matrix A(q) of size l n reflects the l n
velocity constraints imposed on the robot motion. Naturally, in the face of slippage occurrence it is assumed
.
that the velocity constraints A(q)q = 0 are violated.
j
Hence, for A (q) denoting the j -th row of the Pfaf.
fian matrix we arrive at the product sj = Aj (q)q for
j = 1, 2, . . . l defining a slip: either lateral or longitudinal. It will be assumed further that all l phase
constraints are violated.
Assuming that the platform moves on the horizontal plane and its wheels slip and touch the ground
point-wise, the robot dynamics may be represented as
the following control system
..
y = k(q). (1)
.
sL =
Lateral slip
sS =
vR cos vW
vW
vR sin
vW
Driving
vR cos > vW
sL =
vR cos vW
vR cos
sS = tan
(6)
where
Res (sRes ) = c1 (1 exp
(c2 sRes )
) c3 sRes ,
has the meaning of the resultant slip reaction coefficient with parameters c1 , c2 , c3 depending on the
road surface (see Table 2). Is it possible to extend the
Burckhardts approach via two more factors to obtain
[19]
Res (sRes ) = (c1 (1 exp(c2 sRes ) ) c3 sRes )ec4 sRes vCoG
(1 c5 FZ2 ).
(5)
FL sin .
(4)
(7)
(8)
with AiS and AiL denoting the rows of the matrix A(q)
related, respectively, to the lateral and to the longitudinal slip of this wheel. Since each wheel may either
drive or brake, we denote by Fbsi the slip reaction force
for the i-th wheel braking and by Fdsi this force for the
i-th wheel driving. In this way we arrive at 2k admissible variants of motion of the wheeled mobile robot
equipped with k wheels.
Let i = {b, d}, i = 1, 2, . . . , k describes the
motion type (braking/driving) of the i-th wheel. For
all types of slip permitted, the robots motion may be
c1
c2
c3
1.2801
0.857
1.1973
1.3713
0.4004
0.1946
0.05
23.99
33.822
25.168
6.4565
33.7080
94.129
306.39
0.52
0.347
0.5373
0.6691
0.1204
0.0646
0
(9)
P(q) + D(q, q) = Fs (q, q) + B(q)u
y = k(q),
with
.
Fs (q, q) =
k
Fspp .
(10)
p=1
3 Motion Planning
m
gi (x)ui ,
y = k(x),
(11)
(12)
i=1
.
In this section the motion planning problem is formulated. We introduce preliminaries about Jacobian
algorithms derived by means of the ECSA.
For fixed , the motion equations (9) can be represented as a control affine system with output function,
of the form
x = f (x) + g(x)u = f (x) +
du ()
du ()
= Jx0 ,T (u ())
= e(),
d
d
(14)
(15)
is refereed to as the Jacobian of the system (11) at
u(). It describes how an infinitesimal change in the
motion planning algorithm can be adapted to constrained motion planning problems [14], or to the
multiple-task motion planning [24].
3.1 A TaskPriority Motion Planning Algorithm
A(t) =
J#x0 ,T (u())v (t) = BT (t)(T , t)CT (T )G1
x0 ,T (u())v
(18)
(19)
u =0 (t) = u0 (t),
(20)
Kx0 ,T (u()) =
Hi (x(t), u(t))dt,
(22)
(23)
(24)
du (t)
= BT (t)T (T , t)CT (T )G1
x0 ,T (u ())e( ).
d
(21)
where (t) denotes the solution of Eq. 16. The MoorePenrose inverse for 0 Jx0 ,T (u()) is defined by Eq. 18,
bi (t) + ci (t)
v,
||bi (t) + ci (t)||2
(26)
v X , where
bi (t) = B (t)
Hi (x(s), u(s))
(s, t)
x
T
T
ds,
(27)
and
ci (t) =
Hi (x(t), u(t))
u
T
.
(28)
i=1
z
i1
du ()
i
=
j =0
(32)
(30)
> 0 and i = 1, 2, . . . , z. In agreement with [24] the motion planning algorithm for the
i-th subtask takes the form
du (t)
= i i J#x0 ,T (u ())i e() (t)
d
+ i Px0 ,T (u ())i () (t),
(31)
ud (t)
S : y0 yd ,
Sen :
ud (t)
uT (t)udt min
(34)
or
0
ud (t)
S : y0 yd ,
T
ud (t)
1
Sslip :
(A(x1 (t))x2 (t))T A(x1 (t))x2 (t)dt min
0
(35)
Kx0 ,T (u()) =
and
(37)
1J
x0 ,T (u())
Jx0 ,T (u())v() =
T
0
H(x)
(t)dt
x
(38)
(39)
(42)
The implementation of the above task as an additional objective function in the motion planning algorithm results in the control functions characterized by
the minimal energy consumption or by the minimal
slippage effects.
Since the endogenous configuration space is
infinitedimensional, in order to carry out effective computations a finite-dimensional, orthogonal,
trigonometric (Fourier) parameterization of control
functions is employed, by setting
Jx0 ,T () = T
or
Jx0 ,T () =
T
0
H(x(t))
(t)dt,
x
(44)
(45)
= 0, 1, . . . ,
(46)
(47)
where 1 Gx0 ,T () =1 Jx0 ,T ()1 JTx0 ,T (), 0 Px0 ,T () =
I 0 J#x0 ,T ()0 Jx0 ,T () and the task space errors
for both sub-tasks: 0 e() =0 Kx0 ,T ( ) yd and
1 e() =1 K
x0 ,T ( ).
u (t) = Ps (t),
with Ps (t) = diag{P(t), . . . , P(t)} as a block diagonal matrix built of m copies of the row matrix P(t) =
[1, sin t, cos t, . . . , cos pt], = 2/T , containing 2p + 1 basic functions, therefore s = m(2p + 1).
4 Case Study
A Pioneer 2DX platform has been intended to serve as
a tool for implementing and testing a motion planning
(48)
l
Z
(x,y)
r
X
Y
vy
2y
2
y1
vx2
.
x2
.
y
y2
.
x y.
1
r
x2
x1
vx1
.
x1
1
a
0.5
2
1.5
[rad]
y[m]
1
0.5
0.5
0
0.5
1
0
x[m]
c
0.7
1
0
t[s]
0.01
u1
u
0.008
0.5
0.006
0.4
0.004
0.3
0.002
0.2
0
sL []
u[Nm]
0.6
0
0
t[s]
0.05
t[s]
3
2.5
0.04
1.5
1
0.02
s []
0.03
0.5
0
0.01
0.5
0
0.01
0
1
1
t[s]
1.5
0
t[s]
Fig. 3 Motion planning of Pioneer 2DX on the dry asphalt: a platform path X Y , b orientation, c controls, d wheel slips, e wheel
slides, f platform velocities
a
0.5
2
1.5
[rad]
y[m]
1
0.5
0.5
0
0.5
1
0
1
0
x[m]
d
0.7
0.025
u2
s
0.02
0.5
sL []
u[Nm]
0.6
0.01
0.3
0.005
0.1
L2
0
0
t[s]
f
s
2.5
S1
0.08
L1
t[s]
0.015
0.4
0.2
0
2
t[s]
sS2
2
1.5
1
sS []
0.06
0.5
0.04
0
0.02
0.5
1
0
0
t[s]
1.5
0
t[s]
Fig. 4 Motion planning of Pioneer 2DX on the wet asphalt: a platform path X Y , b orientation, c controls, d wheel slips, e wheel
slides, f platform velocities
truncated Fourier series are assumed (chosen) composed of the constant term and up to the 2nd order
harmonics.
We shall proceed according to the following algorithm:
a
b
0.4
2
1.5
1
[rad]
y[m]
0.2
0.2
0.5
0
0.4
0.5
0.6
0
1
0
x[m]
d
0.7
0.65
0.2
2
t[s]
sL1
sL2
0.6
0.15
0.5
s []
0.45
u[Nm]
0.55
0.1
0.4
0.35
0.05
0.3
0.25
0
0
0
t[s]
1.5
t[s]
f
s
S1
2.5
S2
2
1.5
s []
1
0.5
0.5
0
0.5
1
0
0
t[s]
t[s]
Fig. 5 Motion planning of Pioneer 2DX on the wet cobblestones: a platform path X Y , b orientation, c controls, d wheel slips, e
wheel slides, f platform velocities
based on the state x(ti1 ) from the previous time instant decide about the motion type
(braking/driving) and choose a proper subsystem (9) for further calculations (according to
Table 1)
a
0.5
0.4
1.5
0.2
0.1
(d,d)
(b,d)
[rad]
y[m]
0.3
0.5
(b,b)
(b,b)
(d,b)
0.1
0.2
0
0.5
0
x[m]
0.7
sL1
u2
sL2
0.65
0.6
0.2
2
t[s]
0.15
0.5
s []
0.45
u[Nm]
0.55
0.1
0.4
0.35
0.05
0.3
0.25
0
0
0
t[s]
2.5
t[s]
f
2
S1
sS2
1.5
1
s []
S
1.5
0.5
0
0.5
0.5
1
0
0
t[s]
t[s]
Fig. 6 Motion planning of Pioneer 2DX on the snow: a platform path X Y , b orientation, c controls, d wheel slips, e wheel slides,
f platform velocities
1 T T
2 0 u udt
1(a) (0 S)
1(a) (0 S +1 Sen )
1(b) (0 S)
1(b) (0 S +1 Sen )
2(a) (0 S)
2(a) (0 S +1 Sslip )
2(b) (0 S)
2(b) (0 S +1 Sslip )
0.42
0.22
0.51
0.32
5.30
5.24
6.29
6.17
9.05 104
2.66 104
1.70 104
5.17 104
5.19
4.79
3.17 102
2.76 102
T
0
vT vdt
10.39
7.46
10.30
8.60
17.36
17.49
23.56
23.93
T
0
.T .
v vdt
13.06
7.61
2.76
1.74
42.89
43.15
31.49
31.93
a
b
0.4
1.5
0.6
1
[rad]
y [m]
0.2
0.8
1
0.5
0
1.2
0.5
1.4
1.6
0
1
0
x [m]
0.4
0.35
2.5
0.3
3
t [s]
x 10
sL []
u [Nm]
0.25
0.2
1.5
0.15
0.1
0.5
0.05
0
0
0.5
1.5
0.5
0
t [s]
t [s]
x 10
14
s []
12
10
1.5
0.5
4
0
0.5
0
2
0
t [s]
t [s]
Fig. 7 Taskpriority motion planning of Pioneer 2DX (0 S +1 Sen ); yd = (5, 0, 2 ): a platform path X Y , b orientation; c controls,
d wheel slips, e wheel slides, f platform velocities
the wet asphalt. At this point the parking maneuver problem is considered, thus
the robot starts with the initial posture
x0 = (0, 0, l 4 , 0, 0, 0, 104 , 104 , 0, 0, 0)T
(a) In order to observe the slippage minimization effect, we let the robot move on
0.8
1.5
0.6
[rad]
y [m]
2.5
0.4
0.2
0.5
0
0
0
0
t [s]
x [m]
x 10
0.4
2.5
0.35
sL []
u [Nm]
0.3
0.25
0.2
0.15
1.5
1
0.1
0.5
0.05
0
0
0.5
1.5
0.5
0
t [s]
x 10
2.5
1.5
s []
t [s]
0.5
4
5
0
t [s]
0
0
t [s]
Fig. 8 Taskpriority motion planning of Pioneer 2DX (0 S +1 Sen ); yd = (5, 3, 0): a platform path X Y , b orientation; c controls,
d wheel slips, e wheel slides, f platform velocities
a
0.2
0.1
2.5
u [Nm]
y [m]
0.1
0.2
1.5
0.3
0.4
0.5
0
0.5
x [m]
0.2
0.5
t [s]
1.5
1.5
s []
S
sL []
0
1
0.2
0.5
0.4
0
0.5
t [s]
1.5
0
0
0.5
t [s]
1.5
Fig. 9 Taskpriority motion planning of Pioneer 2DX (0 S +1 Sslip ); yd = (5, 0, 2 )): a platform path X Y , b controls, c wheel
slips, d wheel slides
2.5
2.2
2
1.8
1.5
u [Nm]
y [m]
1.6
1.4
u1
u2
u1
u2
1.2
0.5
S + S sli p
S
0
0
x [m]
0.03
0.025
S + 1S sli p
S + 1 S sli p
0
S
0
S
0
0.5
1.5
1.5
t [s]
0.02
sS []
0.01
0.02
s []
:
:
:
:
0.015
0.03
0.01
0.04
0.005
0.05
0
0
0.5
t [s]
1.5
0.5
t [s]
Fig. 10 Taskpriority motion planning of Pioneer 2DX (0 S +1 Sslip ); yd = (5, 3, 0): a platform path X Y , b controls, c wheel slips,
d wheel slides
5 Conclusions
Modeling and control of wheeled mobile robots whose
motion is subject to the longitudinal and lateral slips
of all their wheels has been studied. Depending on
possible wheels slip conditions (i.e. braking/driving),
several modes of motion have been proposed, resulting in the model belonging to a subclass of switched
control affine systems with autonomous switches.
This paper shows applicability of the Endogenous
Configuration Space Approach as a tool for motion
planning of Wheeled Mobile Robots with multifacet
dynamics. Specifically, a taskpriority motion planning algorithm for a Wheeled Mobile Robots, able
to solve the problem composed of two subtasks has
been designed. These subtasks include reaching a
desired point in the task space along with the minimization of either the control effort or the wheels
slippage. Performance of the algorithm has been illustrated by solving a parking maneuver problem of
the Pioneer 2DX robot. Computer simulations have
shown that the control function provided by the algorithm achieves a motion of the robot, from an initial state to a terminal state either with constrained
torques exerted by robots actuators or with less wheels
slippage. The motion planning algorithm based on
the Endogenous Configuration Space Approach has
proved to be able to control the switched robotic system effectively (providing smooth controls and motion
devoid of chattering effects). This paper focuses only
on motion planning that can be open-loop and offline. It delivers just a reference trajectory of the robot.
Tracking control needs to be realized in the closed
loop and on-line. A predictive control algorithm or
a backstepping method are recommended to this
purpose.
The conceptual basis presented in this paper opens
several avenues for further research. The next step
of the presented research will consist in the experimental evaluation of the proposed approach on the
Pioneer platform. Special attention should be paid to
augmenting the motion planning problem with a task
References
1. dAndrea-Novel, B., Campion, G., Bastin, G.: Control of
wheeled mobile robots not satisfying ideal velocity constraints: a singular perturbation approach. Int. J. Robust
Nonlinear Control 5, 243267 (1995)
2. Antonelli, G.: Stability analysis for prioritized closed-loop
inverse kinematic algorithms for redundant robotic systems.
IEEE Trans. Robot. 25, 985994 (2009)
3. Broggi, A., et al.: Intelligent vehicles. In: Springer Handbook of Robotics, pp. 11751198. Springer, Berlin (2008)
4. Burckhardt, M.: Fahrwerktechnik: Radschlupf-Regelsysteme, Wurzburg, Vogel Fachbuch (1993)
5. Caldwell, T.M., Murphey, T.D.: Switching mode generation and optimal estimation with application to skisteering.
Automatica 47, 5064 (2011)
6. Canudas de Wit, C., Tsiotras, P., Velenis, E., Basset,
M., Gissinger, G.: Dynamic friction models for road/tire
longitudinal interaction. Veh. Syst. Dyn. 39(3), 189226
(2003)
7. Chitour, Y., Sussmann, H.J.: Motion planning using the
continuation method. In: Essays on Mathematical Robotics,
pp. 91125. Springer, New York (1998)
8. Cortes Monforte, J.: Geometric, Control and Numerical
Aspects of Nonholonomic Systems. Springer, New York
(2002)
9. Dixon, W.E., Dawson, D.M., Zergeroglu, E., Behal, A.:
Nonlinear Control of Wheeled Mobile Robots. Springer,
Berlin (2001)
10. Giergiel, J., et al.: Symbolic generation of the kinematics equations of the mobile robot Pioneer 2DX. Przeglad
Mechaniczny 1920(/2000), 2631 (2000) (in Polish)
11. Gonzalez, R., Fiacchini, M., Alamo, T., Guzman, J.L.,
Rodriguez, F.: Adaptive control for a mobile robot under
slip conditions using a LMIbased approach, pp. 1251
1256. In: Proceedings of the European Control Conference
(2009)
12. Hendzel, Z., Trojnacki, M.: Neural network identifier of a
fourwheeled mobile robot subject to wheel slip. J. Autom.
Mob. Robot. Intell. Syst. 8(4), 2430 (2014)
13. Iagnemma, K., Ward, Ch.C.: Classificationbased wheel
slip detection and detector fusion for mobile robots on
outdoor terrain. Auton. Robot. 26, 3346 (2009)
14. Janiak, M., Tchon, K.: Constrained motion planning for
nonholonomic systems. Syst. Control Lett. 60, 625631
(2011)
15. Jung, S., Wen, J.T.: Nonlinear model predictive control for
the swingup of a rotary inverted pendulum. J. Dyn. Syst.
Meas. Control 126, 666673 (2004)
16. Karnop, D.: Vehicle Dynamics, Stability, and Control. CRC
Press, Boca Raton (2013)
17. Khan, H., Iqbal, J., Baizid, K., Zielinska, T.: Longitudinal and lateral slip control of autonomous wheeled mobile
robot for trajectory tracking. Front Inform. Technol. Electron. Eng. 16(2), 166172 (2015)
42. Zefran,
M., Burdick, J.W.: Design of switching controllers
for systems with changing dynamics. In: Proceeding of 37th
IEEE Conference on Dec. & Contr., pp. 21132118 (1998)
Katarzyna Zadarnowska received the Ph.D. degree in Technical Sciences in 2005. She has been participated in several research projects, e.g. Endogenous configuration space
approach to mobile manipulators (supported by Polish State
Committee of Scientific Research), Hybrid systems with applications to automotive control (supported by Marii Curie Training Site), Green Transfer academia-to business knowledge
transfer project (co-financed by the European Union under the
European Social Fund) or RobREx - Autonomy in rescue and
exploration robots (supported by the Polish National Centre for
Research and Development). Since 2005 she has been an Assistant Professor at the Institute of Computer Engineering, Control
and Robotics, Wrocaw University of Technology, Poland. Currently she works at the Chair of Cybernetics and Robotics,
Wrocaw University of Technology, Poland. Her current interests include mobile robots, wheel slippage modeling, motion
planning algorithms, hybrid control systems, nonholonomic
systems, optimal control, stochastic algorithms and computational fluid dynamics.