Sie sind auf Seite 1von 6

Trajectory Tracking Control Strategy using Co-Reference for

Rear-Steered Vehicle *
Anugrah K. Pamosoaji

AbstractA trajectory tracking control strategy using a co-

reference for a rear-steered non-holonomic vehicle is presented. v
Given a reference moving on a path, the vehicle has to track the r
reference and reduce the vehicle-to-reference distance to the qr
minimum possible distance. In addition, the vehicle has to reference
maintain the change rate of the vehicle-to-reference distance r
non-positive at the neighborhood of the reference from any v
initial configuration. A class of barrier Lyapunov Function
(BLF) is utilized to analyze the stability of the proposed control P
strategy. The co-reference is used to guide the vehicle to enter vdr
the neighborhood of the reference from the rear side of the q
reference. The performance of the proposed control strategy is actual
evaluated by simulations and the results are presented. vehicle
Keywords: Trajectory tracking control, automated guided
vehicle, distance reduction. Fig. 1. Trajectory tracking scenario. The actual vehicle intends to get closer to
the related reference.

introduced. The co-reference will guide the vehicle such that

I. INTRODUCTION it will get closer to the reference from the rear of the actual
Research on tracking of moving reference has been reference. In this paper, a control algorithm utilizing barrier
almost-mature in the latest years. A large number of tracking Lyapunov function (BLF) is designed to drive the vehicle to
control strategies has been developed. Some works used its respective reference focusing on reducing vehicle-to-
Cartesian coordinate system to express the vehicles reference distance.
configuration [1]-[5]. The methods have a difficulty, i.e., the The contribution of this study is the design of trajectory
vehicle-to-target distance cannot directly be controlled. The tracking control algorithm to reduce the vehicle -to-reference
vehicle-to-target distance is necessary especially for the distance under significantly distinct initial orientation.
problem of tracking moving preplanned references in multiple
Reducing the distance is important, especially when the For navigation purpose, we use navigation variables, as
initial configuration of the vehicle is significantly distinct to shown in Fig. 1.
the planned reference. Suppose that in the warehouse cases, if
the material handling vehicles are failed to get closer to the q(t ) r T . (1)
related preplanned reference, the entire plans of other vehicles
might be violated. To address this problem, a polar coordinate where r , , , and , are the vehicle-
system is recommended to use, since one of its axis can to-reference distance, the inclination angle of the vehicle-to-
represent such the distance. Some trajectory tracking control reference line OP with respect to the vehicles orientation,
strategies using polar coordinate system are reported in [6]- and the inclination of the vehicle-to-reference line OP with
[16]. respect to the references orientation, respectively. The
formulations of the navigation variables are described as
Some problems in using polar coordinate system are follows.
raised when the vehicle enters the environment of the
reference from very large orientation angle, for instance, from ~ ~2
x y , if x, y r ,
the front side of the reference. This situation is not r (2)
advantageous, since the vehicle-to-reference distance will be ~ x2 ~
y 2 , if x, y f ,
increased as a compensation of orientation adjustment in
small distance. arctan 2( ~y , ~x ) q , (3)
To attack such the problems, a new technique of applying
guidance point which takes a role as a co-reference is q g arctan 2( ~y , ~
x) , (4)

where ~ x xr x and ~ y yr y ; r and f represents

*Research supported by Universitas Atma Jaya Yogyakarta. the vehicle is in the rear of its reference and the vehicle is
A. K. Pamosoaji is with the Faculty of Industrial Technology, Universitas
Atma Jaya Yogyakarta, 55281 Indonesia. (phone: +62-274-487-711 ext. in the front of the reference, respectively, and are described
3157; fax: +62-274-485-223; e-mail: as
r x, y / 2, / 2 , (5) vr r 1 sin r . (15)

f x, y , / 2 / 2, . (6) It can be concluded that is influenced by uncontrolled

variables, i.e., vr and , besides the control law vdr , . In
The time derivative of the navigation variables is expressed as
this situation, it is difficult to drive to zero. Meanwhile,
r vr cos vdr cos cos , (7) is fully uncontrollable, according to (15). In other case, if the
1 1 1 initial configuration is in 3 ( r , , ) r r , | | 0 ,
v r r sin vdr r sin cos vdr l sin , (8)
then that the influence of vr and r to can be eliminated.
v r r 1 sin v dr r 1 sin cos r . (9) However, the term r 1 in (26) is still appeared and might
In addition, as shown in Fig. 2, the vehicle attempts to track a cause scattering on , especially under the condition of
preplanned reference moving on a path. The reference can be r 0 and sin r . Therefore, we need to drive to
represented as a point Pr xr , yr moving along a path with zero under a proper value of r such that is bounded. By
linear and angular velocities comparing (14) and (26), it can be concluded that by
achieving | | 0 first following by zeroing and reducing
vr xr2 y r2 (10)
r until r r , we can guarantee that the condition (12)
In addition, the direction of the reference is symbolized as r never occur for all t 0 .
and is formulated as
Another problem exists when we intend to guarantee that
r xr , yr arctan y r , xr . (11) at small vehicle-to-reference distance r , the changing rate of
r is always non-positive or r does not exceed r . From
(7), we can conclude that there are some initial configurations
III. PROBLEM DESCRIPTION that makes r increase in some time interval, i.e., when the
In this study, we focus on a trajectory-tracking problem condition of vdr cos cos vr cos is satisfied.
where the motion of the reference (linear and angular
velocities) is independent to the position of the vehicle. The The objective of the study is to design a control law
reference, whose linear and angular velocities are known, vdr , for the vehicle such that the vehicle can reach

moves along a predefined path. Let 1 ( r , , ) r r , | | , and | | at t T , where T is finite time for any
where r is sufficiently small. Suppose that the vehicle starts initial configuration, and there are some trajectories that
from a configuration ( x0 , y0 ,q 0 ) 1 such that satisfies | | 0 . In addition, the vehicle must avoid collision
( r 0 , 0 ,0 ) 1 . Also, suppose that a control law consisting with any obstacles in the workspace.
a driving velocity vdr 0 and a steering angle
[ / 2, / 2] are applied such that IV. CONTROL DESIGN
A. Transformed Barrier Lyapunov Functions (TBLF)
r 0 and ( x(Ts ) x0 ) 2 ( y (Ts ) y0 ) 2 r 0 , (12)
Consider the following system,
where Ts represents sampling time. It can be investigated z f ( z, u ) , z (0) z0 , (16)
that and will abruptly leap. Assume that there exists a
where z Z n and u U m are the state and control
Lyapunov candidate function such the control law makes the
equilibrium point of the system asymptotically stable. input of the system, respectively; z0 Z is the initial state;
According to (8)-(9), and will swing away from zero f is a piecewise continuous and locally Lipschitz
and the vehicle will rotate rapidly towards 1 . Therefore, it function. Also, define za Z and z b Z , such that za zb
can be concluded that if r 0 and sin r then and z0 ( za , zb ) .
grows larger and make unstable. For the design of u , we introduce a scalar, continuous
Let us focus on the more specific condition. Suppose that function called transformed barrier Lyapunov function
(TBLF) for each zi , symbolized as Vz , i , formulated as

the initial configuration is in 2 ( r , , ) r r , 0 .
Therefore, according to (9), the change rate of and
becomes Proposition 1: Define constants k a,1 0 , k a,2 0 ,
k b,1 0 , k b,2 0 , z 0 . Consider two open regions
r vr cos vdr cos . (13)
D1 {z : ka,1 z kb,1} (17)
vr r 1 sin vdr l 1 sin . (14)
D2 y : ka,2 y kb,2 . (18)

1 q (e z ) k z ,1 ln k b2 k b2 e z2

Let f maps D1 to D2 as

e z f ( z ) z z z , (19) vr e
where z is a translation along z -axis. For any z D1 , R
there exists a e z D2 if and only if z satisfies v r Yr
e e
z z k b,2 z z z k a,2 . (20) j
Proof: Let M be a non-negative constant. Applying er E
z z k b,2 M to (19) yields e z k b,2 M . Since M is O
nonnegative, the maximum ez is ez k b,2 . In other words, if Fig. 2. The vehicle O will move towards guidance point E while E moves
towards the reference R.
M 0 then ez belongs to D1 . On the other hand, applying
z z z ka,2 M to (19) yields ez ka,2 M . Since M Consequently, min V 0 at arg min V z z1 . The
is nonnegative, the minimum ez is ez ka,2 , or in other time-derivative of V (ez ) is
words, if M 0 , e z belongs to D1 , as well.
Definition ddd1: Let us consider a region and
V k z ,1q ( y )e z z k a2 e z2


D2 {ez : ka,2 ez kb,2 } , where ez z z z , z

belongs to D1 {z : ka,1 z kb,1} , and z . A
k z ,1 1 q y e z z k a2 e z2 1
e z

transformed barrier Lyapunov function (TBLF) is a
continuous and positive definite function V (ez ) that is Proposition 2 Consider the TBLF V (e z ) in (23) and the
defined with respect to a system in (16) and has continuous time derivative V in (24). The equilibrium of z 1 z z
first-order partial derivatives on every point in D2 and has a will be asymptotically stable if :
property V (ez ) as ez approaches the boundaries of
D2 , i.e., ez ka,2 and ez k b,2 , and satisfies V (ez ) b for
1) for z z z1 , z k a2 z z z 2 ,
all t 0 along the solution of z f ( z , u ) for z D2 and
some positive constant b .

2) for z z z1 , z k b2 z z z 2 .
Proof: It is obvious that V (ez ) in (23) is positive definite.
Let a barrier Lyapunov function (BLF) is defined as
We need to determine the necessary condition such that V is

V ( z ) 0.5 q ( z )k z ,1 ln k a2 ka2 z 2

negative if z z z1 and zero otherwise. For z z z1 ,
the term z z z is negative. Choosing the upper limit of z

1 q( z ) k z ,1 ln k b2 k b2 z 2 1

is sup z k a2 z z z 2 and applying z sup z yields
V k z ,1 z z z z 0 . For z z z1 , the term
where q(z ) is 1 if z 0 and 0 otherwise.
z z z is positive. Therefore, we choose the lower limit of
The function lies on an open set
D1 {z : ka z k b } . We introduce a transformed

z as z * k b2 z z z 2 . Selecting z z * yields

barrier Lyapunov function (TBLF) defined on the domain V k z ,1 z z z z 0 .

D2 {ez : ka ez k b } , where ez f (z ) and B. Trajectory Tracking Control Design
z , as follows. In this study, we introduce a co-reference directing the
vehicle to the right way such that the vehicle could approach

V (e z ) 0.5 q(e z )k z ,1 ln k a2 k a2 e z2 its reference from the right direction. The main idea is that the
co-reference will take a proper place such that the vehicle

1 q (e z ) k z ,1 ln k b2 k b2 e z2 (22) must get closer to the guide prior to approaching the reference
(see Fig. 2). Suppose that the guide is located at point
It can be investigated that min V 0 at y arg min V 0 . E xe , ye ,q e , where xe , ye and q e are the position and
One can rewrite (22) as the function of z as follows. orientation of the, as shown in Fig. fff5. Let us define the
following parameters. Roughly speaking, there must be a rule
such that the co-reference is always in the rear side of the

V (e z ) 0.5 q(e z )k z ,1 ln k a2 k a2 e z2

reference. In other words, the position of the co-reference
must satisfy
q r arctan 2( yr ye , xr xe ) ( , ) . (25) Note that r , , and are designated as planned r ,
As shown in Fig. fff5, the co-reference E performs new , and at which, according to (40)-(42), the minimum
navigation variables, i.e.m ( r , , ) , where r is the values of Vr , V , and V are located. In other words, r ,
distance between the vehicle to the co-reference, is the , and are designated to be temporary goals for each
inclination angle of the vehicles direction with respect to the associated navigation variables.
vehicle-to-co-reference line, and is the direction of the Proposition 3: Under the free workspace, applying the
co-reference with respect to the vehicle-to-co-reference line. following control law
As the consequence of the co-reference existence, we
vdr k v ,1e r exp(k v, 2 e r ) vr , (34)
formulate new navigation variables er , e , and e as
c h1 h2 l
arctan , (35)
er ye y 2 xe x 2 , (26)

cr h3 1h3 2 2

e arctan 2( ye y, xe x) q , (27) to the co-reference point E with the following guidance rules:
e q g arctan 2( ye y, xe x) . (28) r (t ) r * t , (36)
We construct the relationship between the new navigation r 2 e r2 r2
variables and the original ones as arccos , (37)
2 re r

e r r r r
, (38)
e ( ) , (29)
e ( ) leads the original navigation function to r r , 0 ,

where 0.
Proof: Let us introduce a Lyapunov candidate function
formulated as
z , if z ( , ]
( z) (30) V Vr V V , (39)
z 2sign ( x) , otherwise.
In order to build transformed Lyapunov functions (TBLFs), Where V r , V , and V are candidate TLBFs and are
we define r a and r b , a ( ,0) and described as
b [0, ] , and a ( ,0) and b [0, ] as the lower
and upper limits of r , , and , respectively; and r ,
V (er ) 0.5 q(e r )k r ln r a2 r a2 er2


, and are the equilibrium points of r , , and ,

respectively, and are assumed as constant at a time instance.
1 q (er ) k r ln r b2 r b2 er2 1

Therefore, the changing rate of e r , e , and e with respect
to time can be expressed as follows.
V (e ) 0.5 q(e )k ln a2 a2 e2


er r vr cos vdr cos cos , (31)

e vr r 1 sin

1 q (e ) k ln b2 b2 e2 1


vdr r 1 sin cos vdr l 1 sin , (32)

V (e ) 0.5 q (e )k ln a2 a2 e2

e vr r 1 sin vdr r 1 sin cos r . (33)

To solve the problem of curse of small r , two types of

1 q(e ) k ln b2 b2 e2 1


motions are defined. The first is -adjusting motion, i.e.,

1, if em 0,
the vehicle attempts to reduce to zero while keeping the q ( em ) (43)
distance r sufficiently far to avoid the large altering of . 0, otherwise,
The second one is r -regulating motion. The motion is Furthermore, the substitutions of (7), (8), and (9) to the time
executed under almost-zero and . Problem of scattering derivatives of Vr , V , and V , respectively, yields
appears in this motion, especially when r is closer to zero.

Vr cr r2 vr cos e vdr cos e cos , (44)
v v v
V c 2 r sin e dr sin e cos dr sin , (45) c h1 h2 l
er er l arctan , (57)

cr h3 1h3 2 h22

v v
V c 2 r sin e dr sin e cos r . . (46)
er er

Since Vr , V , and V in (54)-(56) guarantee that the
We design a switching mechanism by applying the following navigation variables are bounded near the equilibrium for
rule for c r , c , and c as follows. any vdr , then we can select any vdr to guarantee that e r 0
for e r 0 . Applying (34) to (31) yields the following

c r q (er ) r a2 er2


(1 q (er )) r b2 er2
k r er ,

(47 necessary condition for e r 0 :

0, if e 0, kv,1er exp(kv,2er )cos cos cos . (58)

c q (e ) a2 e2

k e , otherwise,
(49 Suppose that 0 , 0 . Then the values of k v ,1 and
1 k v , 2 must satisfy
2 2
(1 q (e )) b e

k v ,1e r 1 exp(k v, 2 e r ) . (59)
0, if e 0,

c q (e ) a e
2 2 1

k e , otherwise.

(1 q (e )) b2 e2
A simulation was conducted to evaluate the performance
of the designed control strategy. In the simulation, a reference
Assume that Vr , V , and V are positive definite and the starts from (30 m,10 m, rad) and move to the direction of
time-derivatives are negative semi-definite. Consequently, if x with linear velocity vr 2 m/s . The vehicle starts from
e 0 then V 0 and the Lyapunov candidate function (20 m,20 m, / 2 rad) . In the simulation, the values of k r ,
(39) depends only on Vr and V . It means that the k , and k are set such that c r cos e / v r ,
controllers objective is to decrease e r and e . The situation c tan e / l c r 1 , and c 1 vr sin e / e r r ,
is similar for e 0 . respectively, especially at e r 0 , e 0 , and e 0 . For
Let us define the following parameters: other situation, their values are k r 1 , k 1 , and k 1 .
h1 vr er1 sin e 1 , h2 vr er1 sin e r c , and The resulted trajectory, vehicle- and co-reference- to-
h3 vr cos e . We choose the following equations consisting reference distances, and the orientation navigation angles are
vdr and , i.e. revealed in Figs. 3, 4, and 5, respectively. As shown in Fig 4,
the co-references initial distance to the reference is r * 20

vdr cos e cos c r h3 1 h3 , (51) m and is constantly decreased linearly with respect to time
with 4 . The important result here is that if | | / 2 ,

sin e cos l 1 sin vdr c h1 , (52)
instead of directly moving towards the reference, the vehicle
moves to the co-reference. That is the reason that in Fig 4,
vdr er1 sin e cos h2 , (53) the vehicle-to-reference reduces slowly (from t=3s to t=9s).
to (44), (45), and (46), respectively and we obtain
Vr c r2 r2 vr2 cos 2 e , , (54) A tracking control strategy for moving reference is
presented. The main feature of the proposed method is the
V c2 2 , (55) use of guidance points positioned behind the reference point.
For controller design, a type of Barrier Lyapunov Function
V c2 2 , (56)
(BLF) is used. A simulation is performed and the results
respectively. Therefore, the equilibrium (er , e , e ) (0,0,0) show that the proposed control strategy is stable. The future
works are planned to extends the control strategy to the case
is stable. Eliminating the term vdr er1 sin e cos in (52) and of adversarial workspaces.
(53) and assuming that is in / 2 / 2 gives us the
following formulation for the steering angle: ACKNOWLEDGMENT
The research is supported by the Industrial Electronics
Laboratory, Universitas Atma Jaya Yogyakarta (UAJY),
[3] M. K. Kang, J. S. Lee, and K. L. Han, Kinematic path-tracking of
mobile robot using iterative learning control, Journal of Field
25 Robotics, vol. 22, no. 2, pp. 111-121, January 2005.
[4] M. Malisof, F. Mazenc, and F. Zhang, Stability and Robustness
Analysis for Curve Tracking Control using Input-to-State Stability,
vehicle's initial IEEE Transactins on Automatic Control, vol. 57, no. 5, May 2012.
configuration [5] I. Zohar, A. Ailon, and R. Ravinovici, Mobile Robot Characterized by
15 Dynamic and Kinematic Equations and Actuator Dynamics: Trajectory
Y (m)

Tracking and Related Application, Robotics and Automation Systems,

vol. 59, no. 6, pp. 343-353, 2011.
10 [6] M. Egerstedt, X. Hu, and A. Stotsky, Control of Mobile Platforms
Using a Virtual Vehicle Approach, IEEE Transactions on Automatic
Control, vol. 46, no. 11, pp. 1777-1782, 2001.
reference's initial [7] D. Chwa, Tracking Control of Differential-Drive Wheeled Mobile
configuration Robots Using a Backstepping-Like Feedback Linearization, IEEE
0 Transactions on Systems, Man, and Cybernetics-Part A: Systems and
0 10 20 30 40 Humans, vol. 40, no. 6, pp. 1285-1295, 2010.
X (m) [8] A. Widyotriatmo and K.-S. Hong, Asymptotic Stabilization of
Nonlinear Systems with State Constraints, International Journal of
Fig. 3. Resulted trajectory.
Applied Mathematics and Statistics, vol. 53, no. 3, pp. 10-23, 2015.
[9] A. Widyotriatmo and K.-S. Hong, Switching Algorithm for Robust
20 Configuration Control of A Wheeled Vehicle , Control Engineering
Practice, vol. 20, no. 3, pp. 315-325, 2012.
vehicle-to-reference [10] A. K. Pamosoaji and K.-S. Hong, Time-Constrained Trajectory
distance Tracking Control Algorithm for Multiple-Vehicle Systems, in
Proceedings of 2012 IEEE International Conference on Mechatronics
distance (m)

and Automation, 2012, pp. 1148-1153.

[11] Z. Li, J. Deng, R. Lu, Y. Xu, J. Bai, and C.-Y. Su, Trajectory-
10 Tracking Control of Mobile Robot Systems Incorporating Neural-
co-reference-to-reference Dynamic Optimized Model Predictive Approach, IEEE Transactions
distance on Systems, Man, and Cybernetics: Systems, vol. 46, no. 6, pp. 740-
749, 2015.
[12] E. Kayacan, H. Ramon, and W. Saeys, Robust Trajectory Tracking
Error Model-Based Predictive Control for Unmanned Ground
Vehicles, IEEE/ASME Transactions on Mechatronics, vol. 21, no. 2,
0 pp. 806-814, 2015.
0 5 10 15
[13] A. Widyotriatmo and K.-S. Hong, Configuration control and
Time (s)
autonomous vehicle under nonholonomic and field-of-view
Fig. 4. Vehicle-to-reference distance (solid) and co-reference-to- constraints, International Journal of Imaging and Robotics, vol. 15,
reference distance (dashed). no. 3, pp. 126-139, 2015.
[14] A. K. Pamosoaji, P. T. Cat, and K.-S. Hong, Sliding-mode and
proportional-derivative-type motion control with radial basis function
40 neural network based, estimators for wheeled vehicles, International
Journal of Systems Science, vol. 45, no. 12, pp. 2515-2528, June 2014.
20 [15] A. K. Pamosoaji and K.-S. Hong, Collision-free path and trajectory
0 planning algorithm for multiple-vehicle systems, Proceedings of the
IEEE Conference on Robotics, Automation, and Mechatronics (RAM),
-20 2011, pp. 67-72.
Angles (deg)

[16] K. P. Tee, S. S. Ge, and E. H. Tay. Barrier Lyapunov functions for

-40 the control of output-constrained nonlinear systems, Automatica, vol.
45, no. 4, pp. 918927, 2009.



0 5 10 15

Fig. 5. Resulted and .

[1] A. P. Aguiar, J. P. Hespanha, and P. V. Kokotovic, Path-Following
for Nonminimum Phase Systems Removes Performance Limitations,
IEEE Transactions on Automatic Control, vol. 20, no. 2, pp. 234-239,
[2] T. A., Tamba, B. Hong, and K.-S. Hong, A Path Following Control of
An Unmanned Autonomous Vehicle, International Journal of
Control, Automation, and Systems, vol. 7, no. 1, pp. 113-122, 2009.