2012 20th Mediterranean Conference on Control & Automation (MED)

Barcelona, Spain, July 3-6, 2012

Switching Model Predictive Control for an Articulated Vehicle

Under Varying Slip Angle
Thaker Nayl, George Nikolakopoulos and Thomas Gustfsson

Abstract— In this article a switching model predictive control the effect of the slip angles, as the complexity of the overall
scheme for an articulated vehicle under varying slip angles is problem is being increased and it is more difficult to proceed
being presented. For the non–holonomic articulated vehicle, to the next stage of the control scheme design based on
the non–linear kinematic model that is able to take under
consideration the effect of the slip angles is extracted. This highly non–linear articulated vehicle’s dynamics.
model is transformed into an error dynamics model, which in From a control point of view, there have been proposed
the sequence is linearized around multiple nominal slip angle many traditional techniques for non–holonomic vehicles,
cases. The existence of the slip angles has a significant effect
on the vehicle’s path tracking capability and can significantly based on error dynamics models without the presence of slip
deteriorate the performance of the overall control scheme. angles. More analytically, in [6, 8] linear control feedback
Based on the derived multiple error dynamic models, the has been applied, while in [9] a Lyapunov based approach
varying slip angle is being considered as the switching rule and has been presented. In [10] a control scheme based on
a corresponding switching mode predictive control scheme is LMIs has been presented and in [11] a pole placement
being designed that it is also able to take under consideration: a)
the constrains on the control signals and b) the state constraints. technique has been applied. Moreover, in [12] the authors
Multiple simulation results are being presented that prove the have presented a path tracking controller based on error
efficacy of the overall suggested scheme. dynamics, while in [13] the problem of designing a path
following controller for a n–trailer vehicle, based on non–
I. INTRODUCTION linear adaptive control has been derived. Finally, classical
Among the current vehicle types utilized in a mine field, fundamental problems of motion control for articulated ve-
articulated ones are the most characteristic vehicle’s type that hicles have been presented in [7, 14, 15].
can be found most frequently, such as the Load Haul Dump The main contribution of this article is dual. First, to the
(LHD) vehicles. In general the articulated vehicles consist author’s best knowledge, this is the first time that an error
of two parts, a tractor and a trailer, linked with a rigid free dynamics modeling framework will be derived for the case of
joint. Each body has a single axle and the wheels are all non– an articulated vehicle operating under slip angles. Until now
steerable, while the steering action is performed on the joint, and mainly due to the increasing complexity of the problem,
by changing the corresponding articulated angle, between the only the case of non–slip angles affecting the movement
front and the rear parts of the vehicle. of the vehicle has been considered. Secondly, the problem
During the operation of an articulated vehicle, there are of controling the articulated vehicle, under the presence of
various external factors that degrade the overall system varying slip angles, is being addressed by a switching Model
performance and introduce errors in the model that could be Predictive Control (MPC) scheme.
propagated with the time. Such deteriorating factors are: a) Multiple MPC controllers are being fine tuned for specific
the generic interaction between the vehicle and its surround- slippage operating conditions, while the proposed control
ing environment [1], b) the noise and bias in the positioning scheme has the ability to take under consideration the effect
and driving sensors [2], c) the dynamic effects resulting from of real life constraints on the control input (articulated angle)
acceleration and braking [3], and d) the existence of variation and the environmental restrictions. In this control design
of slip angles among the trailer and the tractor, which are approach the current estimated slip angle of the tractor is
also being influenced by the type of the driving ground, the considered as the mode selector for the switching MPC. The
fatigue and the tyres’ type [4]. resulting control scheme provides the optimal control for
Among the aforementioned factors that degrade the overall each region of slip angles, while ensuring smooth transition
performance of the articulated vehicle, the existence of of the control effort as the articulated vehicle is driven over
slip angles is one of the most significant problem that the regions of different slippage.
modeling and control approaches should face. In the relative The rest of this article is structured as it follows. In
literature there have been several research approaches for the Section II the error dynamics modeling framework for the
problem of modeling articulated vehicles, based on the theory cases of with and without the effect of the slip angles is
of multiple body dynamics [5–7]. Most of these methods being presented. In Section III the switching model predictive
contain simple models that are not taking under consideration controller is being analyzed, while in Section IV multiple
simulation results are being presented that prove the efficacy
The authors are with the Department of Computer, Electrical and Space
Engineering, Luleå University of Technology, SE–97187, Luleå, Sweden of the proposed scheme. Finally the conclusions are draw in
Corresponding Author’s email: [email protected] Section V.

978-1-4673-2531-8/12/$31.00 ©2012 IEEE 890

 II. A RTICULATED V EHICLE M ODELING values when (l f = lr ) or the vehicle is not driving straight
A. Articulated Vehicle Kinematics Without Slip Angles (γ =0).
The case of an articulated vehicle under no slip angles is v f sinγ − l f γ̇ cosγ
θ̇r = (8)
being presented in Figure 1. l f cosγ + lr
B. Articulated Vehicle Kinematics Under Slip Angles
In the case of slip angles, the kinematic model of the
articulated vehicle can be also formulated from general
geometry that includes ideal and slip behavior for the steering
angles. The definition of this model has been initially based
on the derivation in [16], which included the perturbed
factors in the vehicle position as two slip variables β and
α , being defined as the tractor’s and trailer’s slip angles
respectively, and which are also the angles between the linear
velocities of the vehicle’s parts.
For the following derivation of the kinematic model, it
Fig. 1. Typical articulated vehicle without slipping is being assumed that the vehicle is moving on a flat
surface under the influence of slip angles, while the vehicle’s
In this figure (x f , y f ) and (xr , yr ) denote the coordinates configuration is depicted in Figure 2. where (r1 , r2 ) are the
of the front and the rear part of the vehicle, p f and pr are the
corresponding centers of gravity, l f and lr are the length of
the front and the rear units, while the angles θ f and θr denote
the vehicle’s part orientation, and the x, y axes represent the
fixed coordinating system. Moreover v f and vr are the linear
velocities of the tractor and trailer respectively and are being
defined as:
ẋ f = v f cos θ f (1)
ẏ f = v f sin θ f (2)
The articulated angle γ is being defined as the difference
between the orientation angle θ f of the front and the orien-
tation angle θr of the rear part of the vehicle. By examining
the vehicle’s depicted geometry, as also the relation between
Fig. 2. Articulated vehicle modeling configuration under the influence of
the coordinates of p f and pr , the equations of motion for slip angles
the vehicle’s rear part, can be derived from the following
equations: instantaneous centers of velocity for the tractor and the trailer
xr = x f − l f cos θ f − lr cos θr (3) of the vehicle with different radiuses. For the initial center
curvature of the trajectory it is being assumed that the vehicle
yr = y f − l f sin θ f − lr sin θr (4)
is moving forward without slip conditions. In the following
When the non–holonomic constraints, acting on the front and derivation, the unit subscript ’s’ denotes variables in the slip
rear axles, are taken under consideration, mainly resulting case as they have been defined previously for the non–slip
from the assumption of moving without slipping in the one.
vehicle’s wheels, the following equations (5–6) can also be The kinematic equations, for the motion under the effect
extracted: of slip angles, for the front part can be formulated as:

ẋ f sinθ f − ẏ f cosθ f = 0 (5) ẋ f s = v f cos (θ f + β ) (9)

ẋr sinθr − ẏr cosθr = 0 (6) ẏ f s = v f sin (θ f + β ) (10)

By taking the derivative of the equations (3–4) with respect In this case, the vehicle motion depends not only on vehicle’s
to time, substituting the results in equations (5-6), and by speeds, the articulated angle and the vehicle’s lengths, but
further simplifying the results, the time derivative of the also on the slip angles, while the resulting vehicle’s heading
tractor’s orientation angle is being defined as: is provided by a combination of the slip angles with the
orientation angles. The rate of the orientation θ̇rs can be
v f sinγ + lr γ̇
θ̇ f = (7) defined as function of the steering angle γ and both slip
l f cosγ + lr angles. Based on the assumption that the vehicle develops a
The angular velocities for the tractor and the trailer, which steady–state motion turning, this rate can be provided by the
are being defined as θ̇ f and θ̇r respectively, have different utilization of a virtual center of rotation, depending on the

978-1-4673-2531-8/12/$31.00 ©2012 IEEE 891

 velocity as a function of (cos(β ), sin(β )) [2]. With respect
to the local coordinate axis origin, the velocities are being
defined as:
  
cosβ 0 vfs
vfs = (11)
sinβ −l f θ ˙f s
  
cosα 0 vrs
vrs = (12)
sinα lr θ̇ rs

By utilizing rigid body principles, the velocity of the

tractor, with respect to the velocity of the trailer, can be cal-
culated with respect to the articulated angle as it follows [17]:
 −1    
cosα 0 cosγ −sinγ cosβ 0 vfs
vrs = (13)
sinα lr sinγ cosγ sinβ −lr θ ˙f s

By replacing equations (11-12) in (13), the angular speed for

the tractor can be defined as:
v f sin(γ + β − α ) + lr γ̇ cos α
θ̇ f s = (14)
l f cos(γ − α ) + lr cos α
It should be noted that equation (14) is more accurate if the Fig. 3. Articulated vehicle following error transformation
slip angles are known a priori, a task that is quite difficult as
the slip angles are dependent of the vehicle’s speed, mass,
tire–terrain interaction and articulation angle, in a highly θ are small, l can be defined as l = r θ and ed = rθ eh , and
non–linear way. Moreover, in calculating the angular velocity by taking the time derivative of this equation yield:
of the tractor, the primary sources of the errors, are due to ėd = v eh (15)
the time varying parameters γ , γ̇ , β and α , as errors in these
parameters propagate directly to the states. The variables γ The heading error eh is the orientation difference between
and γ̇ can be measured with a great accuracy, while an the centers of the vehicle and the circular path. A change
estimation algorithm or a look up table should be utilized in the heading error is being defined as eh = θ − φ , while
for determining the value of the slip parameters. the relation between θ and φ with respect to l, after small
assumptions, is defined as: (θ r = φ R). From this equation
C. Error Dynamics Modeling it can be derived that eh = θ R− R
θ r
, and by taking the time
Before proceeding with the design of the control scheme, derivative results in: ėh = θ̇ r( r − R1 ). The curvature error ec
1

an appropriate model should be initially derived. In the measures the difference between the vehicle’s path circle and
sequel, the error dynamics modeling procedure will be the curvature of the trajectory path, and it can be defined as:
presented. while it should be highlighted that this is the first ec = ( 1r − R1 ). By utilizing the relations in above equations,
time in the relevant scientific literature that such a modeling the curvature error is being defined as:
framework, with the ability to take under consideration the ėh = v ec (16)
effect of slip angles, is being presented.
Based on the assumption that the vehicle develops a With the assumption that the velocity and the curvature of
the trajectory are constant and by differentiating the curvature
steady–state motion turning, or γ̇ = 0, the rates of the orien- equation with respect to time, the rate of the curvature error
v
tation change are being provided by: r1 = θ̇ f and r2 = θv˙r . In is being defined as:
f r
Figure 3 it is depicted an overview of how the displacement, γ̇ lr cos(γ + β − α )cos α + γ̇ l f cos(γ + β − α ) cos(γ − α )
ėc = + (17)
heading and curvature errors are being defined, between the (lr cosα + l f cos(γ − α ))2
actual path of the articulated vehicle and the desired one. γ̇ v l f sin(γ + β − α ) sin(γ − α ) − γ̇ l f lr cos α ) sin(γ − α )
2
−
v(lr cosα + l f cos(γ − α ))2
In this figure, the distance from the vehicle to the reference
γ̈ l f lr cos(γ − α ) sin(α )α̇ + (lr2 cosα sinα α̇ )
path displacement, as the angle between the vehicle and the v(lr cosα + l f cos(γ − α ))2
reference are also being displayed.
In the following derivation, three errors are defined as: a) Linearizing the error dynamics in equation (17) around the
ed is the displacement error, b) eh is the heading error, and c) reference path γ yields:
ec is the curvature error. Based on [5], the displacement error (l f + lr ) + l f (γ 2 + γβ − 2γα ) + l f (α 2 − αβ )
ed is the difference between the coordinates of the tractor ėc = γ̇ (18)
(lr + l f )2
and the coordinates of the desired circular path. From the
triangle (abd), it can be defined that θ = arctan rl and from where it can be observed that e˙c is dependant on the slippage
the triangle (dbc) eh = arctan( eld ). If it is assumed that φ and and the rate of change of articulation angles. By performing

978-1-4673-2531-8/12/$31.00 ©2012 IEEE 892

 the following change of variable: d) the mode selector signal, applies the necessary switching
optimal control action.
lf (γ 2 + γβ − 2γα )
ė´c = ėc − γ̇ (19)
(lr + l f )2
The following state space error dynamics description for the
articulated vehicle is being extracted:
⎡ ⎤
⎡ ⎤ ⎡ ⎤⎡ ⎤ 0
ėd 0 v 0 ed ⎢ l f (1+β −2α ) ⎥
⎣ėh ⎦ = ⎣0 0 v⎦ ⎣eh ⎦ + ⎢ (l f +lr )2 ⎥ γ̇ (20)
⎣ ⎦
ė´c 0 0 0 éc lr +l f (1+α 2 −α β )
(l f +lr )2

In the case that the articulated vehicle is being driven

over terrains that are characterized by different slip angles,
multiple larger operating sets of slip angles, with the same
characteristics, ai ∈ L1 and bi ∈ L2 , with L1 , L2 ∈ ℜ2 and
i ∈ Z+ , can be defined, leading to a switching state space
description for the model in (20) as it follows:
ẋ = Ax + Bi γ̇ (21)
Fig. 4. Switching MPC scheme block diagram
where:
⎡ ⎤
⎡ ⎤ ⎡ ⎤ 0 The construction of the MP–controller is based on the
ed 0 v 0 ⎢ l f (1+βi −2αi ) ⎥ system description defined in equation (21). The mode
x = ⎣eh ⎦ , A = ⎣0 0 v⎦ , Bi = ⎢
⎣ (l f +lr )2 ⎥
⎦
l +l (1+ α 2 −α β ) selector signal i ∈ S with S  {1, 2, · · · , s} is a finite set
éc 0 0 0 r f i i i
(l f +lr )2 of indexes and s denotes the number of switching sub–
systems in (25). For polytopic description, Σ is the polytope
and x ∈ X ⊆ ℜ3 is the state vector, γ̇ ∈ U ∈ ℜ is the control Σ : Co{[A B1 ], · · · , [A Bs ]}, Co denotes the convex hull
action, A ∈ ℜ6×6 , Bi ∈ ℜ3×1 , and full state feedback is being and [A, Bi ] are the vertices of the convex hull. Any [A, Bi ]
considered, or C = I3×3 . within the convex set Σ is a linear combination of the vertices
∑sj=1 μ j [A Bi ]with ∑ j=1 μ j = 1, 0 ≤ μ j ≤ 1. In the presented
s
III. S WITCHING M ODEL P REDICTIVE C ONTROL D ESIGN
methodology for the design of the MPC scheme, the mode
MPC is a highly effective control scheme that is able
selector signal is the estimated slip angle. For defining
to take under consideration multiplicative system model
the switching instances, the sets L1 and L2 have been
descriptions, uncertainties, nonlinearities and physical and
discretized into equal operating subspaces. The discretization
mechanical constraints in the system model parameters or in
of the the slip angles operating set, can be formulated by
the control signals [18]. In the current research effort, the
defining multiple nominal values α0 , β0 and allowing them
MPC schemes is able to predict future values of the vehicles
to take values into neighboring regions of lengths ξi and ψi .
error dynamics based on the present available information
This can be formulated as:
and the current constraints [19]. The MPC control action,
which is the rate of articulation angle, is based on a finite L1,i = α0,i
min
= α0,i − ξi ≤ α0,i ≤ α0,i
max
+ ξi = α0,i
max
horizon continue time minimization of predicted tracking
L2,i = β0,i
min
= β0,i − ψi ≤ β0,i ≤ β0,i
max
+ ψi = β0,i
max
error with constraints on the control inputs and the state
variables. and by assuming that both slip angles are taking the same
The overall block diagram of the proposed closed loop values at the same time instances, due to the fact that the
system is depicted in Figure 4. For efficiently controlling the length of the articulated vehicle is being considered small,
articulated vehicle, the controller utilizes the current state of results in ξi and ψi and for the switching regions:
motion of the vehicle as well as the next target points of
the reference trajectory. The trajectory planner is generating L1 = L2 = L1,i = L2,i
the desired path, while in the sequel this path (planar
coordinates) are being translated to displacement, heading The sets X and U specify state and input constraints. Let
and curvature coordinates that act as the reference input for the set X contain the x states that satisfy the following
the MPC controller. In the presented methodology for the bounding inequality:
design of the MP–controller scheme, the mode selector signal xmin = x − Δ1 ≤ x ≤ x + Δ1 = xmax
of the MPC is the estimated slip angle for the front part
(3,1)
of the vehicle using estimation approaches (like extended where Δ1 ∈ ℜ+ is the vector containing the selecting state
Kalman filter). The formulation of the MPC is based on: a) boundary conditions. The control input bounding set U can
the current full state feedback, b) the active constrains on be derived by taking under consideration the mechanical and
the system, c) the estimated or measured slip angles, and the physical constraints of the articulated vehicle, as also the

978-1-4673-2531-8/12/$31.00 ©2012 IEEE 893

 preference on aggressive or not maneuvers. These constraints where, n is the index along the prediction horizon, Nw is
can be also formulated as presented in (22). the beginning of the prediction horizon, Q is the output
error weight matrix, R is the rate of change in control
umin = u − Δ2 ≤ u ≤ u + Δ2 = umax
weight matrix, N is the control action error weight matrix,
where Δ2 ∈ ℜ+ is the vector containing the selecting control ŷ(k + n|k) is the predicted system’s output at time k + n,
boundary conditions. Let the matrix Hi be a zeroed 2 × 2 given all measurements up to including those at time k,
matrix with its i–th column equal to [1, −1]T , and the other r(k + n|k) is the output set–point profile at time k + n, given
02,1 i.e. for i = 2: all measurements up to including those at time k, Δu(k + n|k)
  is the predicted rate of change in control action at time
0 1
Hi = k + n, given all measurements up to including those at time
0 −1
k, u(k + n|k) is the predicted optimal control action at time
Then the previous bounds can be cast in a more compact k + n, given all measurements up to and including those at
form as: time k, and s(k + n|k) is the input set–point profile at time
⎡ ⎤ ⎡ max ⎤
  x x + Δ1 k + n, given all measurements up to and including those at
H1
· ⎣−−⎦ ≤ ⎣ − − − − −⎦ time k.
H2 4×2
u 2×1 umax + Δ2 4×1 Once all VMPCi (k) controllers are computed, the total
These constraints are embedded in the Model Predictive switching MP–controller is constructed by implementing a
Control computation algorithm in order to compute an op- switching among the difference controllers in relation with
timal controller that counts for the physical and mechanical the estimated/measured values of slip angles α and β . The
constraints that restrict the articulated vehicle’s motion. objective of the i ∈ S controller is to stabilize the i–th
The basic idea of MPC is to calculate a sequence of future system, while if s increases, then the approximation of the
control actions in such a way that it minimizes a cost function error dynamic modeling is more accurate for a larger part
defined over a predefined prediction horizon. The index of slip angles and thus allowing the development of more
to be optimized is the expectation of a quadratic function efficient flight control algorithms.
measuring the distance between the predicted system’s output IV. S IMULATION R ESULTS
and some predicted reference sequence over the horizon in
addition to a quadratic function measuring the control effort. For simulating the efficacy of the proposed control scheme
More specifically, the (i)–th MP–controller’s objective is to for the problem of path following for an articulated vehi-
optimize the quadratic cost in (22), while the (i)–th linearized cle, over a terrain with varying slip angles, the following
system is within Σ. vehicle’s characteristics have been considered: l f = 0.6m,
Special care must be provided in order to correctly tune the lr = 0.8m and constant speed u = 2m/sec, values that have
prediction N p and the control Nc horizon. A long prediction been extracted from the real articulated vehicle in Figure 1.
horizon increases the predictive ability of the MPC controller Three operating sets for the slip angles have been defined
but on the contrary it decreases the performance and demands as: 0 ≤ |β1 | ≤ 0.02, 0.02 < |β2 | ≤ 0.04, and 0.04 < |β3 | ≤
more computations. The control horizon must also be fine– 0.09, while the bounds on the articulated angle have been
tuned since a short control horizon leads to a controller defined as −0.785 ≤ γ ≤ 0.785, and the bounds on the
that tries to reach the set–point with a few conservative dynamic errors were −0.2 ≤ x ≤ 0.2. The proposed controller
motions, a method that might lead to significant overshoots. has been evaluated in a closed circle reference path. The
On the contrary, a long control horizon produces more corresponding predictive horizon has been set to N p = 10 and
aggressive changes in the control action that tend to lead to the control horizon as Nc = 5 with control interval 0.2sec .
oscillations. Obviously the tuning of prediction and control In Figure 5 the results from path tracking by utilizing
horizon is a coupled process. For the aforementioned reasons switching and non-switching MP–controller under varying
the control horizon must be chosen short enough compared slip angles are being presented. As it can been observed there
to the prediction horizon. Additionally, the response of the is a significant tracking error, due to the effect of slippage
system can be also shaped using weight matrices on the angles deteriorating the overall controller performance.
system outputs, the control action and the control rates. In Figure 6 the error dynamics and in Figure 7 the
Each model predictive controller VMPC i corresponds to the articulation angle and articulation angle’s rate for the two
i-th error dynamic model of the articulated vehicle, obtained cases are being presented. It should be noted that: a) in
by solving the following optimization problem min J(k) with the presented simulation results, harsh switching and big
respect to the control moves variations Δu and to the error slippage angles have been considered, that simulate the
coordinates, k the discrete time sample index and J(k)
defined as: translation of the vehicle under heavy road conditions, and b)
Np the same switching angles have been utilized as the testing
J(k) = ∑ [ŷ(k + n|k) − r(k + n|k)]T Q[ŷ(k + n|k) − r(k + n|k)] + scenario in the depicted simulation results.
n=Nw
Nc −1
+ ∑ [ΔuT (k + n|k)RΔu(k + n|k)] (22) V. C ONCLUSIONS
n=0
Np In this article a switching model predictive control scheme
+ ∑ [u(k + n|k) − s(k + n|k)]T · N[u(k + n|k) − s(k + n|k)]
for an articulated vehicle under varying slip angles has
n=Nw

978-1-4673-2531-8/12/$31.00 ©2012 IEEE 894

 8 been presented. The non-linear kinematic model that is able
Non−Switching MPC
6
Desired path to take under consideration the effect of the slip angles
Switching MPC
was extracted and been transformed into an error dynamics
4
model, which in the sequence has been linearized around
2 multiple nominal slip angle cases. The varying slip angle has
0
been considered as the switching rule and a corresponding
switching mode predictive control scheme was designed. The
−2
simulation results have been presented that prove the efficacy
Y (m )

−4 of the overall suggested scheme. Future work includes the

−6 application of the proposed scheme in experimental studies.
−8 R EFERENCES
−10 [1] J. Apostolopoulos and L. Gordillo, “Numerical comparison in steering
geometry for robotic vehicle by modeling positioning error,” Automa-
−12
tion Robot, vol. 23, pp. 147–159, 2007.
−14
[2] J. Markdahl, G. Bortolin, and U. Andersson, “Traction control for
articulated off-road vehicle,” in Master Thesis, KTH Royal Institute
0 2 4 6 8 10 12 14 16 18 20 22
X (m )
of Technology, 2010.
[3] C. Altafini, “Why to use an articulated vehicle in underground mining
operations?” in International Conference on Robotics and Automation,
Fig. 5. Circle path tracking based on switching MP-controller and non- Detroit, Michigan, May 1999.
switching MP–controller under varying slip angles [4] B. Pacejka, “Tire and vehicle dynamics (2nd ed.),” Society of Auto-
motive Engineers, vol. 23, pp. 147–159, 2005.
[5] P. Ridley and P. Corke, “Autonomous control of an underground
mining vehicle,” in proc. Australian Conference on Robotics and
Displacem ent error

1
Automation, Sydney, Nov. 2001.
0.5
[6] P. Petrov and P. Bigras, “A practical approach to feedback path control
0
for an articulated mining vehicle,” in Pros. of the 2001 IEEE/RSJ, Int.
Conf. on Intelligent Robotics and Systems, aui, Hawaii, USA, Oct.
−0.5
0 50 100 150 200 250 300 350 400 2001.
[7] U. Larsson, G. Zell, K. Hyyppa, and A. Wernesson, “Navigating an
0.5
articulated vehicle and reversing with a trailer,” in IEEE, 1994.
Heading error

[8] P. I. Corke and P. Ridley, “Steering kinematics for a center articulated

0 mobile robot,” IEEE Trans. On Robotics and Automation, vol. 17,
no. 2, 2001.
−0.5 [9] D. Piyabongkarn, R. Rajamani, J. Grogg, and J. Lew, “Development
0 50 100 150 200 250 300 350 400
and experimental evaluation of a slip angle estimator for vehicle
0.5 stability control,” IEEE Trans on Control System Technology, vol. 17,
Curvature error

no. 1, pp. 78–88, January 2009.

0 [10] P. Bigras, P. Petrov, and T. Wong, “A LMI approach to feedback path
control for an articulated mining vehicle,” in Electronics Research,
−0.5
2002.
0 50 100 150 200
Time
250 300 350 400 [11] P. Ridley and P. Corke, “Load haul dump vehicle kinematic and
control,” in Transaction of the ASME, 2003.
[12] A. Hemami and V. Polotski, “Path tracking control problem formula-
Fig. 6. Time evolution of the system error dynamics for both case studies tion of an LHD loader,” International Journal of Robotics Research,
vol. 18, no. 5, 1999.
[13] C. Altafini, “A path tracking criterion for an lhd articulated vehicle,”
in International Journal of Robotics Research, May 1999.
Articulated angle

1
[14] R. M. Densantis, “Modeling and path tracking for a load haul dump
0 mining vehiclel,” Transaction of the ASME, vol. 119, 1997.
[15] P. Bolzern and A. Locatelli, “A comparative study of different solutions
−1
0 50 100 150 200 250 300 350 400 to the path-tracking problem for an articulated vehicle,” in Conference
1 on Control Applications, Glasgow, Scotland, U.K, 2002, pp. 18–20.
Angle rate

[16] S. Scheding, G. issanayake, E. Nebot, and H. Durrant, “Slip Modeling

0
and Aided Inertial Navigation of an LHD,” in International conference
−1 on robotics and automation, New Mexico, 1997.
0 50 100 150 200 250 300 350 400
[17] J. Markdahl, “Traction control for off-road articulated vehicle,” Ph.D.
Switching m o de

3
dissertation, KTH, Jan 2010.
2 [18] F. Nunez, S. Navarro, A. Aguado, and A. Cipriano, “State estimation
based model predictive control for LHD vehicle,” in 17th World
1
0 50 100 150 200 250 300 350 400 Congress, the International Federation of Automatic Control, Seoul,
0.1 Korea, 2008.
Slip angle

[19] F. Kuhne, J. Manoel, and W. F. Lages, “Mobile robot trajectory

0.05
tracking using model predictive control,” in IEEE LARS, 2005.
0
0 50 100 150 200 250 300 350 400
Time

Fig. 7. The articulated angle as a control signal, switching mode selector

and the slip angle

978-1-4673-2531-8/12/$31.00 ©2012 IEEE 895