1 Copyright © 2013 by ASME
Proceedings of the ASME 2013 International Design Engineering Technical Conference & Computers and Information in Engineering Conference
IDETC/CIE 2013 August 4-7, 2013, Portland, Oregon, USA
DETC2013-12401
KINEMATIC MODELING, ANALYSIS AND CONTROL OF HIGHLY RECONFIGURABLE ARTICULATED WHEELED VEHICLES
ABSTRACT
The Articulated Wheeled Vehicle (AWV) paradigm
examines a class of wheeled vehicles where the chassis is
connected via articulated chains to a set of ground-contact
wheels. Actively- or passively- controlled articulations can help
alter wheel placement with respect to chassis during locomotion,
endowing the vehicle with significant reconfigurability and
redundancy. The ensuing ‘leg-wheeled’ systems exploit these
capabilities to realize significant advantages (improved stability,
obstacle surmounting capability, enhanced robustness) over both
traditional wheeled-and/or legged-systems in a range of uneven-
terrain locomotion applications. In our previous work, we
exploited the reconfiguration capabilities of a planar AWR to
achieve internal shape regulation, secondary to a trajectory-
following task. In this work, we extend these capabilities to the
full 3D case – in order to utilize the full potential of kinematic-
and actuation- redundancy to enhance rough-terrain locomotion.
INTRODUCTION Articulated Wheeled Vehicles (AWVs) consist of a principal
vehicle-chassis connected to a set of wheels with ground contact
via actively- or passively-articulated chains. The presence of the
so-called ‘articulated leg-wheel’ chain endows reconfigurability,
by allowing relocation of the wheel with respect to the chassis.
The AWV paradigm offers immense possibilities for
enhanced locomotion-performance of autonomous mobile
robots while assuring the reliability of the system. For instance,
the articulated-suspensions permit the vehicle to change the
location of center of mass by adjusting the linkages/joints, so as
to avoid the rollover when passing the uneven terrain [1, 2].
Similarly, the vehicle can reduce its footprint to pass narrow
doorways or expand the wheelbase when stability is required.
The resulting class of Leg-Wheel Vehicles (as the AWVs are also
called) have many potential applications (ranging from planetary
exploration [3, 4], agriculture [5, 6] to military and rescue
operations) due to their advantages over traditional wheeled
systems.
Figure 1: Schematic of Leg-Wheel Vehicles.
In most applications, the wheels of an AWR are considered
to be rigid discs with a single point-of-contact with the terrain.
The wheel velocities are governed by a set of non-holonomic
constraints, which permit rolling along the disk-plane without
allowing lateral side-slip. However, potentially, these velocity-
level constraints can be violated, but results in slipping and
skidding. Minimization of slipping and skidding is usually
desired (both from the perspective of reducing the energy
dissipation and improving measurement uncertainty) and can be
achieved by adding intermediate articulations. Additionally, the
articulations and actuation within the leg-wheel chain, serves to
Aliakbar Alamdari Mechanical and Aerospace
Engineering,
SUNY at Buffalo Buffalo, NY, 14260
Xiaobo Zhou Mechanical and Aerospace
Engineering,
SUNY at Buffalo Buffalo, NY, 14260
Venkat N. Krovi Mechanical and Aerospace
Engineering, SUNY at Buffalo
Buffalo, NY, 14260 [email protected]
2 Copyright © 2013 by ASME
redirects the motion and forces from axle prior to their reaching
chassis serving as an effective suspension. Thus, the choices for
the topology, dimension and location of articulations (and
subsequent actuation) affects the overall performance of the
vehicle.
Proper design and control schemes for the articulated-leg-
wheel system are needed to achieve proper coordination of the
rolling and steering of the wheels and serves to motivate our
efforts. There is a clear need to quantitatively examine the role
of topologies, dimensions, and configurations of wheels-
articulation within the individual sub-chain as well as the overall
attachment of the sub-chains to the central chassis. In our
previous work [7, 8], we explored the role of that design
selection process by developing systematic analysis and
evaluating performance of the planar version AWR. In this paper,
we extend it into the full 3D case, focusing on exploiting the
kinematic- and actuation-redundancy for reconfigurability of the
system while enhancing the stability of the vehicle in rough
terrain locomotion.
The rest of the paper is organized as follows: In Section 2
we briefly survey the background and literature on other AWRs.
Section 3 briefly examines the development of a computational
approach for twist-based kinematic modeling followed by the
kinematic modeling of a Wheel Leg Actively Articulated Vehicle
(WLAAV) as a concrete case study in Section 4. In Section 5, a
kinematic control scheme is developed for realizing a primary
trajectory-following task while secondarily ensuring constant
roll-pitch by the WLAAV chassis. Finally, Section 6 studies the
performance via simulations followed by some remarks on
ongoing work in Section 7.
BACKGROUND The main research in passive AWRs concerns designing
suspension mechanism to negotiate with uneven terrain. They
change their configuration according terrain topology. Passive
AWRs such as Shrimp [4], Sojourner [10] and Nomad [11] are
designed to have fewer DOF such that the weight of the system
can be supported by the structure. The main advantages of
passive AWRs are in terms of power consumption, payload
capacity and controller design.
Actively articulated vehicles enhance the mobility of the
robots to obtain better performance such as stability and traction,
as demonstrated by Sample Return Rovers (SRR) [12], Athlete
[13], Workpartner [14], Hylos [15] and Azimut [16]. On the other
hand, more actuators, extra weight and control complexity will
be added to the system.
Kinematic modeling of ordinary wheeled mobile robots
[17] as well as classification scheme [18] have been explored.
Tarokh et al. [19] explored kinematic modeling, analysis and
balance control for high mobility wheeled rover traversing
bumpy terrain and illustrated using the highly articulated multi-
task rover (MTR) example.
Similarly, Grand et al. [20] proposed a general kinetostatic
formulation for articulated wheeled rovers moving on uneven
terrain. Their method was applied to the motion control of
wheeled-legged rover based on the decoupling of the posture and
trajectory parameters.
The estimation of wheel terrain contact angles based on
extended Kalman filter for improving ground traction and to
reduce power consumption was investigated in [21]. Two
innovative approaches were presented in [22] for wheeled-
legged vehicle, by defining adhesion and stability coefficients as
traversability indicators. The authors also propose an innovative
optimal force distribution method to make the front and rear
wheel adhesion coefficients equal. Jarrault et al. [23] explored
the contact stability optimization of robust obstacles passage for
high mobility wheeled legged robot. Their optimization
algorithm used both kinematic redundancies to modify the
position of center of mass of rover and distributing of contact
forces, and the actuation redundancy to improve the frictional
contacts.
In our own work, we focus on developing a systematic and
general-purpose modeling, analysis and operational framework,
suitable for both the design and control of such articulated
wheeled vehicles. We build upon our previous work [7, 8] and
focus on extending it to aid the design, analysis and control of a
fully 3D vehicle traversing significantly rough uneven terrain.
COMPUTATIONAL TWIST-BASED KINEMATIC MODELING
Figure 2 depicts a general model of the wheeled-legged
actively articulated vehicles (WLAAV). We define the inertial
frame of reference {F} = (Of,X,Y,Z), the robot platform has a
frame {B}= (Ob,bx,by,bz) attached to the body at the center-of-
mass of platform Ob. This allows us to easily examine both
translational , ,b b bx y z and rotational velocities , ,b b b of
the frame {B} with respect to the inertial frame ( b is the roll,
b is pitch, and b is yaw angle).
A general vehicle chassis is assumed to possess ‘n’ leg-
wheel branches, each with an arbitrary number of articulations
and ending with disk-wheel. The {W}= (Ow, wx,wy,wz),
coordinate frame of the wheel, is attached to the wheel axle, and B
WA is the homogenous transformation that allows coordinates
in the {W} to be expressed in the {B} frame. In our modeling,
each wheel is assumed to be represented by a rigid disc in contact
with a non-deformable terrain, i.e. wheel-terrain contact
mechanics are considered in this simplified form in order to
facilitate development of the control scheme.
3 Copyright © 2013 by ASME
Figure 2: General articulated wheel robot.
The last frame in each leg-wheel chain is the wheel terrain
contact frame {C}= (Oc,Cx,Cy,Cz), with its x-axis is tangent to
the terrain at the point of contact and y-axis is normal to the
terrain.
x z zC C W and y x zC C C
is contact angle defined as the angle between xC and zW
(where xC always lies in the wheel plane). This angle can be
measured using force sensor on the wheel axle, or could be
estimated by Kalman filtering [12]. This angle is considered
constant and equal to zero for flat surface moving. Following the
formulation conventions from Murray et al. [24], the twist matrix
of frame B w.r.t frame A (and expressed in frame A) is given by:
1[ ] A
A A A
B B BT A A (1)
and can re-express the twist matrices/vectors in any convenient
frame of reference via the Adjoint transformation. The velocity
of contact frame in frame F can be obtained by:
1 ...F F F F F
F B m W F
B W C CV V V V V (2)
The Adjoint transformationC
F can be easily written by given
the homogenous transformation ( , )C C C
F F FA R r .
ˆ
0
C F O C
F FC
F C
F
R r R
R
(3)
ˆF Or is a 3 3 skew symmetric matrix, and allows for twist
vectors to be transformed from frame {F} to frame {C}. Using
the chain rule for homogenous transformations (and
differentiating it), one can express the twist of the contact point
respect to the fixed frame in the contact frame of reference as: 1
1 1 ...C B
F C F C B
C B B
W CC m W
W W C
V V V
V V
(4)
Individual joint twist can now be re-expressed as:
1 1 16 6 6 1
C B CB C B B
BV V t (5)
Allowing equation (4) to be written in compact form as:
C F
F C F
C F CV V Bq (6)
where, 1 2
2 3[ ... ]C C C C
m W
W CB t t t t is twist
assembled matrix and is a vector consists of m joint variables
through the chain and wheel rotational velocity . Various
contact constraints can be easily imposed in the contact frame of
reference. E.g. If pure rolling condition is assumed to be true, the
constraints at the contact point can be represented as:
4 6 6 1[0 0 0 0]
CT F T
CS V (7)
where 4 4
6 4
2 2
[ ]0
IS
is a wrench basis selector matrix. The
matrix S represents the direction where force can be exerted, and
selects the first four rows of twist vector and restricts the
translation motion at the contact point. Substituting Eq. (7) into
(6), the constraint conditions for pure rolling of wheel of
branches can be expressed as:
i
BT C F
B B
P
S V Bq (8)
or
[ ] B
F
i BP V Aq (9)
Assembling of all constraints for all legs of AWV, we could
assemble the kinematics equation from equation (9).
1
2 1
.
.
0
0
BF
B
n
n
P
P A
V q
A
P
(10)
This kinematic model is very useful for a variety of reasons:
First, it forms the basis for the kinematic control of the AWV
while avoiding the slip. Note that the contact velocity C
F
BV
has been eliminated from the equation. Second, the static model
can also be easily extracted, which now relates the contact force
act on the wheel to the torques exerted at the joints and the total
other wrenches applied to the robot.
CASE-STUDY: THE WHEELED-LEGGED ACTIVELY ARTICULATED VEHICLE (WLAAV)
4 Copyright © 2013 by ASME
Figure 3: Kinematic modeling of the WLAAV.
The WLAAV has three leg-wheel sub-chains attached
symmetrically to a triangular platform. Each leg wheel system
features a spatial pantograph mechanism and has five DOF. Four
of these articulations of spatial pantograph have motors attached.
We now perform the kinematic analysis of a highly mobile
maneuverable and reconfigurable wheeled legged actively
articulated vehicle shown in Figure 3. One motor powers the
wheel; one motor for steering of caster and one motor for
changing the elevation of the link (a2) with respect to the
platform and another one attached to the link (a1) where can
change the geometry of the leg.
The linear spring/damper in each leg plays an important
role and acts like a shock absorber and from ground and platform
weight carrier in a parallelogram consisting of four bars. In this
part, we assign coordinate frames, as shown in Figure 4, and set
up the homogenous transformation between the various frames
of references.
2
2
3
1 0
1 0
0 0 2 0
0 0 2 0
( ) in( ) 0 ( ) ( )
in( ) ( ) 0 ( ) ( ),
0 0 1 0
0 0 0 1
( ) 0 ( ) ( )
( ) 0 ( ) ( )
0 1 0 0
0 0 0 1
B
C
C
C
Cos S a Cos a Cos
S Cos a Sin a SinA
Cos Sin a Cos
Sin Cos a SinA
4
4
1 0 3 0
1 1 3 03
( ) ( ) 0 ( )
( ) ( ) 0 ( ),
0 0 1 0
0 0 0 1
0 1 0
( ) 0 ( ) ( )
( ) 0 ( ) ( )
0 0 0 1
C
C
c c cC
A
c c c
Cos Sin a Cos
Sin Cos a SinA
m
Cos Sin nCosA
Sin Cos nSin
( ) 0 ( ) ( )
( ) 0 ( ) ( )
0 1 0 0
0 0 0 1
A
C
Cos Sin rCos
Sin Cos rSinA
(11)
Figure 4: Parameters of wheel-leg.
Twist vectors expressed in local frames are found as:
[ , , , , , ]B
F T
B b b b b b bV x y z (12)
2
2 1[0 0 0 0 ]C
B T
CV a 3
2
3
43
4
4
2 0 0
3 1 1
[0 0 0 0]
[0 0 0 0 ]
[0 0 0 0]
[ 0 0 0 0]
CC T
C
CC T
C
AC T
A c c
CA T
C
V a
V a
V n
V r
Contact twist expressed in contact frame is: 32
2
2 2 2 3
43 4
3 4
CCC BCF C F C B C
C B B C C C C
C A CC CC C A
C C A A C
V V V V
V V V
(13)
Applying the non-holonomic constraints to the system using
equation (7) we get:
1
2 1
.
.
0
0
BF
B
n
n
P
P A
V q
A
P
(14)
Constraints at contact points restrict the motion of the
components of the system such that the cooperation of the
actuators is not straight forward and redundancy in the system
further complicates the control. One can systematically deal with
the overall effect of all constraints on the motion by using our
twist based modeling approach, and then a controller could be
designed by putting the kinematic model into different forms.
KINEMATIC CONTROL OF WLAAV We have formed the kinematic equations of WLAAV in the
previous part. The purpose of this spatial articulated vehicle
design is to enable the robot to change the arrangement of its legs
while moving around on uneven terrain.
Thus there are two control tasks, the first one is path
following in Cartesian space (x, y, z), and the second one is
5 Copyright © 2013 by ASME
posture regulation for controlling the platform’s Pitch, Roll and
Yaw. The robot has 15 internal articulations, with 12 actuated
ones, and the rest passive. Three joints are activated whenever
the changing of elevation of the chassis is required; otherwise
these three joints are turned off. The robot possesses both
kinematic redundancy and actuation redundancy that needs to be
resolved.
The basic control law we will use is close loop resolved
motion rate control (RMRC). In RMRC, Jacobian matrix is used
to map the desired velocity in task space to the velocity in joint
space. And certain velocity controllers are implemented to track
the joint velocity signal. The task space variables are (
, , , , ,b b b b b bx y z ), and controlled variable are ( 1 1 1 1
0, , , ,c
2 2 2 2 3 3 3 3
0 0, , , , , , ,c c ). In this control scheme, we will use
the augmented kinematics method to resolve the redundancy.
Note that bX is desired base motion expressed in inertial
frame, in order to use the inverse kinematic model we transform
into body frame by a rotational matrix.
, , ,
0 0 1 0 0
0 0 1 0 0
0 0 1 0 0
b b b
B
F z y z
b b b b
b b b b
b b b b
R R R R
C S C S
S C C S
S C S C
(15)
This control method is a more general method that could be
potentially used for other AWRs. As there are two control tasks
and the path following part has the first priority, we could use
pseudo inverse and potential function techniques to resolve
redundancy. Recall that the kinematic equation is:
bJX Aq (16)
We first find desired base motion d
bX .Then, to resolve the
kinematic redundancy; we use the following control scheme. # #( )[ ( )]
ph
d d
b b b
q q
q J A X K X X I J J z (17)
where # 1( ) T TJ J JJ is pseudo inverse of J. The overall control
scheme is shown in Figure 5. The particular solution ( pq )
enables the AWV to achieve the primary task (path following by
chassis). The homogeneous part ( hq ) now serves to achieve the
secondary task (regulation of leg configuration).
Figure 5: Redundancy resolution control scheme.
SIMULATION RESULTS The simulation is done in MATLAB with the WLAAV
parameters shown in Table 1. Note that all three leg-wheel sub-
chains are assumed to be identical and symmetrically located
around the chassis.
Table 1: Simulation parameters of the WLAAV
0 160a mm 45m mm
1 35a mm 30n mm
2 110a mm 30r mm
3 110a mm 0 0 0
1 2 30 , 120 , 240
In this simulation, the desired vehicle path and velocity are given
and applied to the system and are the primary requirement while
the posture controller is the secondary criterion. The path
following task is to drive the vehicle to track a circular path with
center at (0, 0) and radius 1000mm. the time trajectories are
given by ( )d
bX Rcos t a and ( )d
bY R sin t b where R is the
radius and ‘a’ and ‘b’ are the coordinates of the center of the
circle. This sinusoidal trajectory will effectively change the
configuration of the legs and wheels. Furthermore, the desired
path is changing sinusoidal in z direction 5 (5 )d
bZ sin t . It is
noted that the leg angles change in such a way to balance the
vehicle.
The control performance of control scheme is shown in
Figure 6. We could see clearly that the vehicle could follow the
desired path smoothly. This result indicates the capability of the
WLAAV for various applications that require the configuration
to change its shape to avoid obstacles and improve stability. The
redundancy problem in the system has been resolved.The error
between the actual and desired paths is shown in the figure which
can be further reduced by using a tuned PID controller.
Figure 6: Simulation for path following.
-1000-500
0500
1000
-1000
-500
0
500
1000
-50
0
50
X (mm)
Trajectory Tracking in Sinusoidal Circular Path
Y (mm)
Z (
mm
)
Desired Path
Actual path
6 Copyright © 2013 by ASME
DISCUSSION The reconfigurability and redundancy of leg-wheel
articulated vehicles creates significant benefits but these
capabilities need to be done by careful modeling, analysis and
control. In this paper, we developed a generic twist-based
kinematic modeling methodology for spatial articulated wheeled
vehicles to help systematize the modeling of such complicated
systems for following analysis and control. The model is a very
general one and could be applied for any articulated system with
active suspensions. The model is used to control the posture of
the robot and its static stability. Results show the validity of the
model and the feasibility of this approach. The main feature of
the work is its generality, e.g. dealing with both active (actuated)
and passive (compliant) joints and linkages. This framework was
deployed in the case-study of a wheeled actively articulated
reconfigurable vehicle capable of such reconfiguration. The
kinematic control scheme are deployed and the results are
studied to solve the issues of maintaining kinematic consistency
of the constraints and resolving the redundancies inherent in such
articulated wheeled robots.
ACKNOWLEDGMENTS This work was supported in part by the National Science
Foundation Award CNS-1314484.
REFERENCES [1] S. Sreenivasan and K. Waldron, "Displacement analysis of
an actively articulated vehicle configuration with extensions
to motion planning on uneven terrain," Transactions of the
ASME. Journal of Mechanical Design, vol. 118, pp. 312–
317, 1996.
[2] K. Iagnemma and S. Dubowski, "Traction Control of wheel
mobile robots in rough terrain with applications to planetary
rovers," International Journal of Robotics Research vol. 23,
pp. 1029–1040, 2003.
[3] H. Hacot, "Analysis and traction control of a rocker-bogie
planetary rover," MS, ME, MIT, 1998.
[4] R. Siegwart, P. Lamon, T. Estier, M. Lauria, and R. Piguet,
"Innovative design for wheeled locomotion in rough
terrain," Robotics and Autonomous Systems, vol. 40, pp.
151-162, 8/31/ 2002.
[5] A. J. Baerveldt, "Agricultural Robotics, in: Autonomous
Robots," 2002.
[6] A. Halme, I. Leppanen, J. Soumela, S. Ylonen, and I.
Kettunen, "WorkPartner: interactive human-like service
robot for outdoor applications," International Journal of
Robotics Research vol. 22, pp. 627-640, 2003.
[7] Q. Fu and V. Krovi, "Articulated Wheeled Robots:
Exploiting Reconfigurability And Redundancy," presented
at the ASME Dynamic Systems and Control Conference,
Ann Arbor, Michigan, 2008.
[8] Q. Fu, X. Zhou, and V. Krovi, "The Reconfigurable
Omnidirectional Articulated Mobile Robot (ROAMeR)," in
12th International Symposium on Experimental Robotics,
Delhi, India, 2010.
[9] R. Volpe, J. Balaram, T. Ohm, R. Ivlev, and P. Rocky,
"Rocky 7: a next generation mars rover prototype," Journal
of Advanced Robotics, vol. 11, pp. 341–358, 1997.
[10] K. Iagnemma and S. Dubowsky, Mobile Robots in Rough
Terrain: Estimation, Motion Planning, and Control with
Application to Planetary Rovers: Springer, 2004.
[11] E. Rollins, J. Luntz, A. Foessel, B. Shamah, and W.
Whittaker, "Nomad: a demonstration of the transforming
chassis," in Robotics and Automation, 1998. Proceedings.
1998 IEEE International Conference on, 1998, pp. 611-617
vol.1.
[12] K. Iagnemma, A. Rzepniewski, S. Dubowsky, and P.
Schenker, "Control of Robotic Vehicles with Actively
Articulated Suspensions in Rough Terrain " Autonomous
Robots vol. 14, pp. 5-16, 2003.
[13] K. K. Hauser, T. Bretl, J.-C. Latombe, and B. Wilcox,
"Motion Planning for a Six-Legged Lunar Robot," in In
proceeding of: Algorithmic Foundation of Robotics, New
York, NY, USA, 2006.
[14] S. Ylönen and A. Halme, "Further development and testing
of the hybrid locomotion of WorkPartner robot," presented
at the Proceedings of 5th International Conference on
Climbing and Walking Robots, Pariisi Ranska, 2002.
[15] C. Grand, F. BenAmar, F. Plumet, and P. Bidaud, "Stability
and Traction Optimization of a Reconfigurable Wheel-
Legged Robot," The International Journal of Robotics
Research vol. 23, pp. 1041–1058, 2004.
[16] F. Michaud, D. Letourneau, M. Arsenault, Y. Bergeron, and
R. Cadrin, "Multi-modal locomotion robotic platform using
leg–track–wheel articulations," Autonomous Robots, vol.
18, pp. 137–156, 2005.
[17] P. F. Muir and P .Neuman, "Kinematic modeling of wheeled
mobile robots," Journal of Robotic Systems, vol. 4, pp. 281-
340, 1987.
[18] G. Campion, G. Bastin, and B. Dandrea-Novel, "Structural
properties and classification of kinematic and dynamic
models of wheeled mobile robots," Robotics and
Automation, IEEE Transactions on, vol. 12, pp. 47-62,
1996.
[19] M. Tarokh, H. D. Ho, and A. Bouloubasis, "Systematic
kinematics analysis and balance control of high mobility
rovers over rough terrain," Robotics and Autonomous
Systems, vol. 61, pp. 13-24, 1// 2013.
[20] C. Grand, F. Benamar, and F. Plumet, "Motion kinematics
analysis of wheeled–legged rover over 3D surface with
posture adaptation," Mechanism and Machine Theory, vol.
45, pp. 477-495, 3// 2010.
[21] K. Iagnemma and S. Dubowsky, "Traction Control of
Wheeled Robotic Vehicles in Rough Terrain with
Application to Planetary Rovers," I. J. Robotic Res, pp.
1029-1040, 2004.
[22] S. Nakamura, M. Faragalli, N. Mizukami, I. Nakatani, Y.
Kunii, and T. Kubota, "Wheeled robot with movable center
of mass for traversing over rough terrain," in Intelligent
Robots and Systems, 2007. IROS 2007. IEEE/RSJ
International Conference on, 2007, pp. 1228-1233.
7 Copyright © 2013 by ASME
[23] P. Jarrault, C. Grand, and P. Bidaud, "Robust obstacle
crossing of a wheel-legged mobile robot using minimax
force distribution and self-reconfiguration," in Intelligent
Robots and Systems (IROS), 2011 IEEE/RSJ International
Conference on, 2011, pp. 2753-2758.
[24] R. Murray, Z. Li, and S. Sastry, A mathematical introduction
to robotic manipulator, 1994.