Vous êtes sur la page 1sur 11

OPTIMAL CONTROL APPLICATIONS AND METHODS

Optim. Control Appl. Meth. 2013; 34:191201


Published online 14 February 2012 in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/oca.2016

Minimum-time anti-swing motion planning of cranes using


linear programming
J. J. Da Cruz1, * , and F. Leonardi2
2

1 Escola Politcnica da USP, So Paulo, SP, Brazil, CEP 05508-900


Centro Universitrio da FEI, So Bernardo do Campo, SP, Brazil CEP 09850-901

SUMMARY
A novel method to solve the minimum-time anti-swing motion planning problem for cranes based on the
use of linear programming is proposed. It is shown that its solution can be obtained by solving a sequence
of fixed-time maximum-range linear programming problems. A convergence proof is presented. A classical
kinematical model for which the trolley acceleration is the control variable is used. The crane motion equations are discretized in time assuming that the control variable is piecewise constant. Inequality constraints
on both trolley speed and acceleration are included to represent physical bounds associated to the electromechanical driving system. The load and the trolley are required to be at rest both at the initial and at the final
times. The load cable length as a function of time is assumed given. The ease of both the formulation and
the solution of the problem is in contrast with the traditional two-point boundary value problem associated
to the Pontryagins minimum principle. Copyright 2012 John Wiley & Sons, Ltd.
Received 18 January 2011; Revised 8 December 2011; Accepted 10 December 2011
KEY WORDS:

motion planning; minimum-time control; anti-swing control; crane control; linear


programming

1. INTRODUCTION
Figure 1 depicts a typical crane used for ship unloading operations. Such a crane can be viewed
as a trolleypendulum system, where the pendulum length can be increased or decreased with time
independently of the trolleys motion. A grab located at the end of the suspension cables is used to
grasp and hold the load to be moved. All motions are assumed to occur in the plane of the figure.
Both the initial and final points of the path are given where both the trolley and the pendulum must
be at rest.
In this paper, a novel method to solve the minimum-time anti-swing control problem of cranes by
using linear programming (LP) is proposed. It is shown that its solution can be obtained by solving
a sequence of fixed-time maximum-range problems. A convergence proof is presented.
Because the goal here is to propose a motion planning scheme for cranes, only the generation of
reference trajectories is of concern. A closed-loop control system is required so that a real crane can
follow the reference trajectory under the action of external disturbances and model uncertainties.
This is exactly the same point of view taken in [1].
The problem of anti-swing motion planning of crane systems has been largely studied in the literature for a long time [18]. A variety of approaches has been proposed to deal with the problem
depending on the mathematical model adopted, on the boundary conditions to be satisfied, and on
the performance index to be minimized.

*Correspondence to: J. J. Da Cruz, Escola Politcnica da USP, So Paulo, SP, Brazil, CEP 05508-900.
E-mail: jaime@lac.usp.br
Copyright 2012 John Wiley & Sons, Ltd.

192

J. J. DA CRUZ AND F. LEONARDI

TROLLEY
CABLE

GRAB

HOPPER

PIER
SHIP

Figure 1. Scheme of a crane used for ship unloading.

In [7], the authors consider the load swing minimization of a container transfer subject to a set
of specified boundary conditions. An algorithm for computing the optimal control is proposed by
solving a two-point boundary value problem (TPBVP).
Reference [3] contains an approach to solve the minimum-time transfer of a hoisting load using
the Pontryagins minimum principle. Both the trolley and the hoisting motions are obtained subject
to the condition that the load must be at rest both at the beginning and at the end of the motion.
Mechanical and electrical subsystems are modeled in detail.
The computation of time-optimal trajectories for tower cranes is considered in [4]. A mathematical model is derived in discrete-time form, and computation of sub-optimal paths is performed under
control and state constraints by using a sequential quadratic programming method. A good initial
estimate of the minimum value of travel time is needed.
In [8], the use of reinforcement learning for the computation of time-optimal anti-swing control
is described. The motion was divided into six subtasks and an optimization problem was defined
for each of them. A reinforcement learning algorithm together with a multilayer perceptron neural
network as a value function approximator were used in the optimization.
A lot of work has been carried out on problems of vibration reduction for flexible structures
described by linear time-invariant (LTI) models with parametric uncertainties [914]. In particular,
reference [9] is quite complete and contains an extensive discussion of several related problems.
In [12], the equivalence of minimum-time zero vibration input shaping with the traditional timeoptimal control problem is studied. It is shown that several input shapers can be recast into the
traditional time-optimal control problem of a system subject to actuator bounds. Lau and Pao [15]
used the KarushKuhnTucker optimality condition to prove that minimum-time input shaping is
equivalent to time-optimal control. In both articles, LTI models of flexible structures are considered.
Because in practice the hoisting trajectories are frequently generated independently of the trolley
motions [16], the load cable length as a function of time will be assumed as given in this work.
As a consequence, the crane can be described by a linear time-varying dynamical model. This is in
contrast with the previously cited papers dealing with flexible structures whose models are all LTI.
The trolley acceleration is chosen as the control variable. The crane dynamics are initially modeled as a system of differential equations, which are then discretized assuming that the control
variable is piecewise constant. Constraints on both the trolley speed and the trolley acceleration are
included in the problem to represent physical bounds associated to the electromechanical driving
system. As it was previously said, both the initial and final points of the load path are given where it
is required that both the trolley and the load must be at rest. The goal is to obtain a minimum-time
control law for the crane motion.
Copyright 2012 John Wiley & Sons, Ltd.

Optim. Control Appl. Meth. 2013; 34:191201


DOI: 10.1002/oca

MINIMUM-TIME ANTI-SWING MOTION PLANNING OF CRANES USING LINEAR PROGRAMMING

193

Figure 2. Scheme of the trolleypendulum system.

It is well known that the search for solutions of constrained optimal control problems is generally
a hard task that involves the use of the Pontryagins minimum principle to formulate a nonlinear
TPBVP [9, 17, 18]. The central idea behind the solution proposed here is to set the value of the final
time and then maximize the mean value of the trolley speedthis is the key point to formulate the
problem by using LP because the functional turns out to be linear in this case. The value of the final
time is then iteratively adjusted until the distance traveled by the trolley is close to the goal within a
given tolerance.
One of the remarkable advantages of formulating the problem as an LP one is that LP algorithms
always find the global optimal solution if it exists. With the proposed approach, it is also quite simple to deal with equality and inequality linear constraints on state and control variables. These issues
are in opposition to the majority of TPBVP algorithms, which, in addition, usually require a good
initial guess in some neighborhood of the solutionLP algorithms generally do not even require
that an initial guess be given by the user.
Data from a ship crane currently operating at Sepetiba Port, Brazil, were used to illustrate the
application of the proposed method. Comparison with a direct collocation method is discussed.
The remainder of the paper is organized as follows. Section 2 describes the crane mathematical
model adopted. Section 3 presents the proposed approach to solve the minimum-time problem by
means of a sequence of fixed-time maximum-range problems. Section 4 contains the discretization of such a maximum-range problem and its LP formulation. Results obtained are presented in
Section 5. Finally, Section 6 summarizes the main points of the paper.
Throughout the paper, an over-hat indicates variables expressed in engineering units; the only
O which is measured in radians, and , which has been normalized. The absence of
exceptions are ,
the over-hat indicates dimensionless variables. The derivatives with respect to dimensionless time
are denoted by over dots.
2. THE MATHEMATICAL MODEL
The continuous-time model presented in this section is a classical one that follows closely
reference [3].
It is assumed that the trolley electric drive system is such that a given acceleration uO can be
imposed to the trolley. In other words, uO can be chosen as the control variable. As a consequence,
the system dynamics are independent of the load mass. Additionally, the following hypotheses are
made: the crane is assumed infinitely rigid, dissipative effects are neglected , the load is a point

Linear viscous friction could be easily included in the model.

Copyright 2012 John Wiley & Sons, Ltd.

Optim. Control Appl. Meth. 2013; 34:191201


DOI: 10.1002/oca

194

J. J. DA CRUZ AND F. LEONARDI

mass, the rope is massless and inextensible, and the influence of hoisting acceleration on load swinging is neglected. It should be recalled that such hypotheses are usual in this context and have been
made by several authors [25, 7, 19, 20].
The crane can be modeled as the trolleypendulum system sketched in Figure 2.
The following set of dimensionless variables is defined:
uO
,
uO max

uD

tD

q
g=
O O min tO,

D

(2)

O gO
,
uO max O min

(3)

gO O
,

(4)

D

uO max

D

vD

(1)

O
O min

q
vO g=
O O min
uO max

(5)

(6)

where tO is time, uO D d2 O .tO/=dtO2 is the trolley acceleration, uO max is the maximum trolley acceleration, gO is the magnitude of the local gravity acceleration, O is the current length of the load cable,
O min is the minimum length of the cable along the unloading cycle, O is the trolley position with
respect to a chosen reference, O is the load cable angle with respect to vertical position, and vO D PO
is the trolley speed.
For usual working conditions and absence of load swing at the motion boundaries, the load cable
angle with respect to the vertical is small. As a consequence, the system dynamical equations can
be linearized and written as follows [3]:
R D u

(7)

R C 2P P C  D u,

(8)

where  is assumed here as being a given function of time. This system of equations is clearly
linear time varying. This fact is essential for the development of the LP approach described in the
following.
Notice that this model contains no parametric uncertainties [16].
3. THE MINIMUM-TIME PROBLEM
In this section, two propositions that are the bases for the use of LP to solve the minimum-time
problem are proved.
With no loss of generality, the following initial condition is adopted:
 .0/ D 0.

(9)

Both the trolley and the load are assumed at rest at t D 0:


P .0/ D 0,
Copyright 2012 John Wiley & Sons, Ltd.

(10)
Optim. Control Appl. Meth. 2013; 34:191201
DOI: 10.1002/oca

MINIMUM-TIME ANTI-SWING MOTION PLANNING OF CRANES USING LINEAR PROGRAMMING

195

.0/ D 0,

(11)

P
.0/
D 0.

(12)

Assuming that at the final time t D tf the position f of the trolley is given and both the trolley and
the load must be at rest, one can write
 .tf / D f ,

(13)

P .tf / D 0,

(14)

.tf / D 0,

(15)

P f / D 0.
.t

(16)

For all t 2 0, tf , equipment limitations require that


j u.t / j6 1,

(17)

j P .t / j6 P max ,

(18)

where P max is the given maximum trolley speed.


In the minimum-time problem, the goal is to minimize tf with respect to u, that is,
min tf .

(19)

The minimum-time problem given by Equations (7)(19) will be denoted by PT .f /.


In the following discussion, it is shown that the solution to the minimum-time problem can be
obtained by solving a sequence of fixed-time maximum-range problems to be defined. Thus, assume
that tf is given and consider the problem of maximizing the functional
max  .tf /

(20)

subject to the constraints given by Equations (7)(12) and (14)(18). Such maximum-range problem
will be denoted by PR .tf / and the optimal value of the functional by   .tf /. Clearly,   .tf / > 0
for all tf > 0.
Proposition 1
For a given tf , assume that u solves PR .tf / and   .tf / is the corresponding optimal value of the
functional (20). If   .tf / D f , then u solves PT .f / and the optimal value of the functional (19)
is tf D tf .
Proof
For any admissible u, one has
 .tf / D v.0,
N tf /.tf ,

(21)

where v.0,
N tf / is the mean value of the trolley speed on the interval 0, tf .
In particular, if u D u solves the PR .tf /, then
  .tf / D vN  .0, tf /.tf ,

(22)

where vN .0, tf / is the maximum mean speed on the interval 0, tf  because tf is fixed in PR .tf /
and   .tf / is maximal.
Assume now that it happens that this value of   .tf / turns out to be equal to the given f .
Because vN  .0, tf / is maximal and f is given, from Equation (22) it follows that tf is minimal (i.e.,
tf D tf ). Since u is admissible for PR .tf /, it also solves PT .f /. This concludes the proof. 


In case the maximum trolley acceleration and deceleration are different, this condition can be easily modified.

Copyright 2012 John Wiley & Sons, Ltd.

Optim. Control Appl. Meth. 2013; 34:191201


DOI: 10.1002/oca

196

J. J. DA CRUZ AND F. LEONARDI

Proposition 2
If tf 2 > tf 1 , then   .tf 2 / >   .tf 1 /.
Proof
Let u1 and u2 be the solutions, respectively, to PR .tf 1 / and PR .tf 2  tf 1 /, and let   .tf 1 / and
  .tf 2  tf 1 / be the corresponding optimal values of the functional (20). For t 2 0, tf 2 , define
u0 .t / as

u1 .t /
if t 6 tf 1
(23)
u0 .t / D
u2 .t  tf 1 / if tf 1 < t 6 tf 2 .
The constraints of Equations (9)(18) are obviously satisfied for such an u0 .t / because at t D tf 1 ,
both the trolley and the load are at rest. If  0 denotes the corresponding trolley total displacement,
then
 0 D   .tf 1 / C   .tf 2  tf 1 /.

(24)

  .tf 2  tf 1 / > 0

(25)

  .tf 2 / >  0 ,

(26)

Since

and

the proof is concluded.

As a consequence of Propositions 1 and 2, the following algorithm solves PT .f / within a given
tolerance " > 0:
initialize tf
solve PR .tf /
while j   .tf /  f j> "
if   .tf / < f then increase tf
if   .tf / > f then decrease tf
solve PR .tf /
end
tf

tf

An interval where tf can be searched may be defined as follows. A lower bound of it is given
immediately by f =vmax ; an upper bound can be obtained, for instance, by successively doubling
the lower bound until a value of tf is obtained such that   .tf / > f . A binary search algorithm is
a good choice to select the new intervals of search.
4. SOLUTION OF PR
In this section, the problem PR .tf / will be discretized and written in the form of an LP problem.
It is assumed that u.t / is stepwise constant. From a practical point of view, this is a reasonable
hypothesis if a sufficiently large n is chosen and the time-step is taken as
t D

tf
D ti  ti 1 .1 6 i 6 n/,
n

(27)

where t0 D 0 and tn D tf .
Copyright 2012 John Wiley & Sons, Ltd.

Optim. Control Appl. Meth. 2013; 34:191201


DOI: 10.1002/oca

MINIMUM-TIME ANTI-SWING MOTION PLANNING OF CRANES USING LINEAR PROGRAMMING

197

In a slight abuse of notation, u.i/ is used to denote the value of u.t / for all t in the ith time interval ti 1 , ti /. Taking into account that the initial conditions are null (Equations (9) and (10)) and
considering that u.t / is stepwise constant, the solution to Equation (7) can be written immediately
by integrating it as
P .k/ D t

k
X

u.i/

(28)

i D1

 .k/ D

k
.t /2 X
2.k  i/ C 1u.i/
2

(29)

i D1

for k > 0, where P .k/ and  .k/ respectively denote the values of P and  at time t D kt .
Equation (20) thus leads to the objective function
max  .n/ D
u

n
.t /2 X
2.n  i/ C 1u.i/.
2

(30)

i D1

In view of Equation (28), conditions (14) and (18) can be rewritten respectively as
n
X

u.i/ D 0

(31)

i D1

and
 P max 6 t

k
X

u.i/ 6 P max

(32)

i D1

for 1 6 k 6 n.
The next step is to obtain the discrete form of conditions (15) and (16). Define the state vector
P /0 .
x.t / D .t / .t

(33)

Denoting x.k/ D x.tk / and taking into account that x.0/ D 0 (Equations (11) and (12)), the solution
to Equation (8) can be written as [21]:
x.k/ D

k
X

 k .i/u.i/,

(34)

i D1

where
Z

ti

 k .i/ D

.tk , t /b.t / dt ,

(35)

ti 1

"

b.t / D

(36)
1=.t /

and .., ./ is the state-transition matrix associated to


2
3
0
1
5.
A.t / D 4
P /=.t /
1=.t / 2.t
Copyright 2012 John Wiley & Sons, Ltd.

(37)

Optim. Control Appl. Meth. 2013; 34:191201


DOI: 10.1002/oca

198

J. J. DA CRUZ AND F. LEONARDI

In view of Equations (33) and (34), conditions (15) and (16) can be rewritten as
n
X

 n .i/u.i/ D 0.

(38)

i D1

Finally, condition (17) takes the form


 1 6 u.k/ 6 1 .1 6 k 6 n/.

(39)

Since the objective function (30) as well as the conditions (31), (32), (38), and (39) are all linear in
u.k/, .1 6 k 6 n/, PR .tf / has thus the form of an LP problem.
5. RESULTS
The proposed approach has been used to define the controller currently running at the Sepetiba Port
cranes. To cope with both the variations of tide height and load condition of the ship, an optimal
solution to each of the various initial cable lengths was computed and stored; in the same way,
solutions corresponding to several distances from the ship to the hopper were also previously evaluated and stored. At operation time, a lookup table procedure is then used to search for the solution,
which is closest to the actual values of initial cable length and distance of interest, being thus unnecessary to solve the optimal control problem in real-time; such solution is then given as the reference
signal to be followed by the DC motor servo that controls the trolley motion. These features thus
characterize a typical optimal control motion planning scheme.
The application of the proposed method is illustrated in what follows. The following set of parameters of the Sepetiba cranes were used in this case: PO max D 2.4 m/s, O min D 16 m, O max D 30 m,
uO max D 1.0 m/s2 , and a constant speed of 2 m/s for load hoisting/lowering. A constant load hoisting/lowering speed is a reasonable assumption in this case because the time required to accelerate
the load between 0 and 2 m/s is negligible. The horizontal distance from the ship to the hopper is
O f D 46.1 m.
In the spirit of what has been said in the first paragraph of this section, the results presented in
what follows are model based and hence generated by numerical simulation.
The minimum value obtained for the travel time was 24.3 s.
In Figure 3(a) and (b), the cable length and the trolley acceleration are shown as functions of
time. The trolley acceleration satisfied the constraint uO 6 1.0 m/s2 . It can also be seen that it
resembles a bang-0-bang pattern, which is typical for minimum-time problems, where not only
the control but also a state variable reach their maximum magnitudes to produce a constrained
minimum-time motion.
30
1

Trolley acceleration (m/s2)

Cable length (m)

28
26
24
22
20
18
16
14
0

0.8
0.6
0.4
0.2
0
0.2
0.4
0.6
0.8
1

10

15

20

25

10

15

Time (s)

Time (s)

(a) Cable length

(b) Acceleration

20

25

Figure 3. Cable length and trolley acceleration versus time.


Copyright 2012 John Wiley & Sons, Ltd.

Optim. Control Appl. Meth. 2013; 34:191201


DOI: 10.1002/oca

MINIMUM-TIME ANTI-SWING MOTION PLANNING OF CRANES USING LINEAR PROGRAMMING

199

50

2.5

45
40

Trolley position (m)

Trolley speed (m/s)

1.5

35
30
25
20
15
10

0.5

5
0
0

10

15

20

0
0

25

10

15

Time (s)

Time (s)

(a) Speed

(b) Position

20

25

20

25

Figure 4. Trolley speed and position versus time.


8

Angular speed (o/s)

Angle (o)

2
0
2
4
6

0
2
4
6

8
10
0

10

15

20

25

8
0

10

15

Time (s)

Time (s)

(a) Angle

(b) Angular speed

Figure 5. Load cable angle and angular speed versus time.

The trolley speed is shown in Figure 4(a), where it can be seen that the trolley is at rest both at the
beginning and at the end of the motion. As expected, the constraint PO 6 2.4 m/s has been satisfied.
Figure 4(b) shows the position of the trolley along time. It can be noticed that the total distance
traveled is 46.1 m as desired.
Figure 5(a) and (b) show, respectively, the angle and the angular speed of the load cable as functions of time. As can be seen, the pendulum is at rest at the vertical position both at the beginning
and at the end of the motion.
Because both the trolley and the load are at rest at the end of the motion, they will remain at this
state for later times.
A direct collocation method [22] has also been applied to solve the problem using the nonlinear dynamical model derived by Auernig and Troger [3]. Results obtained agree closely with those
obtained by the proposed method for the example given previouslythe plots were visually indistinguishable from each other. Nevertheless, the direct collocation method has some drawbacks: (i)
the number of variables is large because the problem unknowns are constituted by both the control and the state variables; (ii) a good initial guess is required to start the search process for the
optimal solution; (iii) there is no guarantee that the global minimum will be achieved; and (iv) the
computation of partial derivatives of the objective function as well as of the constraints with respect
to the problem unknowns is required. The tests performed indicate that the LP method runs about
Copyright 2012 John Wiley & Sons, Ltd.

Optim. Control Appl. Meth. 2013; 34:191201


DOI: 10.1002/oca

200

J. J. DA CRUZ AND F. LEONARDI

seven times faster than the direct collocation method when the gradients are computed analytically
and the Hessian is evaluated numerically.
6. CONCLUSIONS
This paper has shown how to perform minimum-time anti-swing motion planning for cranes by
using LP. A simple approach has been proposed where equality and inequality linear constraints on
state and control variables can be dealt with ease. It has been shown that the solution to the problem
can be found by solving a sequence of fixed-time maximum-range LP problems. Convergence of
the method has been proven.
It is well known that LP algorithms always find the global optimal solution of a problem whenever
it exists and the choice of an initial guess is not a critical issue for themfor simplex, in particular,
the user has not to be concerned about it. On the other hand, traditional TPBVP optimal control
formulations based on the Pontryagins minimum principle usually require a good initial guess in
some neighborhood of the solution and are usually hard to solve. In addition, direct search methods
such as direct collocation have some drawbacks such as those mentioned in the previous section.
Hence, the LP proposed approach seems to be an interesting alternative to solve the minimum-time
anti-swing control problem of cranes.
It should be mentioned that as long as the dynamical equations of the motor and trolley are linear,
the proposed approach can be extended to solve a problem including them in the model with essentially no additional difficulty. It could also be easily extended to include bounds on the load angle
with respect to the vertical position.
The proposed approach has been used to define the controller currently running at the cranes of
the Sepetiba Port, Brazil. A mean reduction of about 10% in the crane traveling time was obtained in
the real plant. This can be considered a significant improvement in the system performance because
the crane does not have a feedback signal from the cable angle. Hence, there was no compensation
for model uncertainties and disturbances affecting the load motion.
Results obtained in laboratory tests using a cartpendulum kit with a constant-length pendulum
agree closely with those given by the method.
A natural extension of this work would be to consider the length of the load cable not as a given
function of time but as an unknown one that must be determined to reduce the traveling time. This
point is the object of a paper under preparation. The presence of uncertainties on such function is
also the object of current research.
ACKNOWLEDGEMENT

The first author is grateful to CNPq for its partial support to this research (grant no. 303202/2009-2).

REFERENCES
1. Lee HH. Motion planning for three-dimensional overhead cranes with high-speed load hoisting. International
Journal of Control 2005; 78(12):875886.
2. Al-Garni AZ, Moustafa KAF, Nizami SSAKJ. Optimal control of overhead cranes. Control Engineering Practice
1995; 3(4):12771284.
3. Auernig JW, Troger H. Time optimal control of overhead cranes with hoisting of the load. Automatica 1987;
23(4):437447.
4. Golafshani AR, Aplevich JD. Computation of time-optimal trajectories for tower cranes. Proceedings of the IEEE
Conference on Control Applications, 1995; 11341139.
5. Liang YC, Koh KK. Concise anti-swing approach for fuzzy crane control. Electronics Letters 1997; 33(2):167168.
6. Mita T, Kanai T. Optimal control of the crane system using the maximum speed of the trolley (in Japanese with
English abstract). Transactions of the Society of Instrument and Control Engineering 1979; 15:833838.
7. Sakawa Y, Shindo Y. Optimal control of container cranes. Automatica 1982; 18(3):257266.
8. Scardua LA, Da Cruz JJ, Costa AHR. Optimal control of ship unloaders using reinforcement learning. Artificial
Intelligence in Engineering 2003; 16(3):217227.
9. Dhanda A, Franklin GF. Optimal control formulations of vibration reduction problems. IEEE Transactions on
Automatic Control 2010; 55(2):378394.
Copyright 2012 John Wiley & Sons, Ltd.

Optim. Control Appl. Meth. 2013; 34:191201


DOI: 10.1002/oca

MINIMUM-TIME ANTI-SWING MOTION PLANNING OF CRANES USING LINEAR PROGRAMMING

201

10. Dhanda A, Franklin GF. Equivalent representations of vibration reduction problems. Proceedings of the American
Control Conference, 2007; 49995004.
11. Liu Q, Wie B. Robust time-optimal control of uncertain flexible spacecraft. Journal of Guidance, Control and
Dynamics 1992; 15(3):597604.
12. Pao LY, Singhose WE. On the equivalence of minimum time input shaping with traditional time-optimal control.
Proceedings of the IEEE Conference on Control Applications, 1995; 11201125.
13. Singer NC, Seering WP. Preshaping command inputs to reduce system vibration. Journal of Dynamic Systems,
Measurement and Control 1990; 112:7682.
14. Singhose W, Derezinski S, Singer N. Extra-insensitive input shapers for controlling flexible spacecraft. Journal of
Guidance, Control and Dynamics 1996; 19(2):385391.
15. Lau MA, Pao LY. Comparison of input shaping and time-optimal control of flexible structures. Proceedings of the
American Control Conference, 2001; 14851490.
16. Lee HH. A new motion-planning scheme for overhead cranes with high-speed hoisting. Journal of Dynamic Systems,
Measurements and Control 2004; 126(2):359364.
17. Kamien MI, Schwartz NL. Dynamic Optimization. Elsevier North Holland: Amsterdam, 1981.
18. Seiersted A, Sydsaeter K. Sufficient conditions in optimal control theory. International Economic Review 1977;
18(2):367391.
19. Benhidjeb A, Gissinger GI. Fuzzy control of an overhead crane performance comparison with classic control. Control
Engineering Practice 1995; 3(12):16871696.
20. Strip DR. Swing-free transport of suspended objects: a general treatment. IEEE Transactions on Robotics and
Automation 1989; 5(2):234236.
21. Franklin GF, Powell JD, Workman ML. Digital Control of Dynamic Systems. Addison-Wesley: Reading, MA, 1990.
22. Betts JT. Practical Methods for Optimal Control Using Nonlinear Programming. SIAM: Philadelphia, PA, 2001.

Copyright 2012 John Wiley & Sons, Ltd.

Optim. Control Appl. Meth. 2013; 34:191201


DOI: 10.1002/oca

Vous aimerez peut-être aussi