Applicable Analysis and Discrete Mathematics

available online at http://pefmath.etf.rs

Appl. Anal. Discrete Math. 12 (2018), 192–204.

https://doi.org/10.2298/AADM170426003A

LINEAR PROGRAMMING PROBLEMS

ON TIME SCALES

Rasheed Al-Salih and Martin Bohner∗

In this work, we study linear programming problems on time scales. This
approach unifies discrete and continuous linear programming models and ex-
tends them to other cases “in between.” After a brief introduction to time
scales, we formulate the primal as well as the dual time scales linear program-
ming models. Next, we establish and prove the weak duality theorem and
the optimality conditions theorem for arbitrary time scales, while the strong
duality theorem is established for isolated time scales. Finally, examples are
given in order to illustrate the effectiveness of the presented results.

1. INTRODUCTION

It is well known that discrete-time linear programming problems have numer-

ous applications in areas such as portfolio optimization, crew scheduling, manufac-
turing, transportation, telecommunication, agriculture, and so on. Continuous-time
linear programming problems were first studied by Bellman [3] as a bottleneck pro-
cess. He established the weak duality theorem and optimality conditions. A com-
putational approach has been presented by Bellman and Dreyfus [4]. The strong
duality theorem was studied by Tyndall [27,28] and Levinson [24]. Grinold [21] has
established strong duality without discretizing the continuous problem. A numer-
ical solution to continuous-time linear programming was considered by Buie and
Abrham [19]. Wen, Lur, and Lai [29] have presented an approximation approach
to solve continuous-time problems.
∗
Corresponding author. Martin Bohner
2010 Mathematics Subject Classification. 34N05, 90C05, 90C39.
Keywords and Phrases. Time scales, primal and dual problem, weak duality theorem,
optimality condition, strong duality theorem.

192
 Linear Programming Problems on Time Scales 193

The theory of time scales, on the other hand, was first introduced by Stefan
Hilger in 1988 in his PhD dissertation, see [22]. The purpose of this theory is to unify
discrete and continuous analysis and to offer an extension to cases “in between”.
Many applications in mathematical modelling exist for this theory, e.g., to optimal
control [2, 15–18, 25], population biology [6], calculus of variations [5, 7, 10], and
economics [1, 8, 11, 12, 26].
In this work, it is therefore meaningful to consider linear programming prob-
lems on arbitrary time scales T as an extension of the continuous-time linear pro-
gramming problems and to extend the results in [21, 24, 27–29] to arbitrary time
scale. The paper is organized as follows. In Section 2, some examples related to
time scales calculus are given. In Section 3, the basic structures of the primal lin-
ear programming model as well as the dual model on time scales are formulated.
The weak duality theorem and the optimality condition theorem are presented and
proved in Section 4. Section 5 states and proves the strong duality theorem on
isolated time scales. Examples are presented in Section 6 in order to demonstrate
our theoretical results. In Section 7, some conclusions are given.

2. TIME SCALES CALCULUS

In this section, instead of introducing the basic definitions, derivative, and

integral on time scales, we refer the reader to the monographs [9, 13, 14], in which
comprehensive details and complete proofs are given. For readers not familiar with
the time scales calculus, we give the following few examples. Throughout, T is
the time scale, σ is the forward jump operator, µ is the graininess, f : T → R is
a function, f σ = f ◦ σ is the advance of f , f ∆ is the delta derivative of f , and
Rb
a
f (t)∆t is the time scales integral of f between a, b ∈ T.
Example 2.1. If T = R, then

σ(t) = t, µ(t) ≡ 0, f ∆ (t) = f 0 (t) for t ∈ T,

and Z b Z b
f (t)∆t = f (t)dt, where a, b ∈ T with a < b,
a a
is the usual Riemann integral from calculus.
Example 2.2. If T = {tk ∈ R : k ∈ N0 } with tk < tk+1 for all k ∈ N0 consists only
of isolated points, then

f (tk+1 ) − f (tk )
σ(tk ) = tk+1 , µ(tk ) = tk+1 − tk , f ∆ (tk ) = for k ∈ N0 ,
tk+1 − tk
and
Z tn n−1
X
(2.1) f (t)∆t = µ(tk )f (tk ), where m, n ∈ N0 with m < n.
tm k=m
194 Rasheed Al-Salih and Martin Bohner

The examples in Section 6 are specific cases of Example 2.2 as follows.

Example 2.3. Let h > 0. If T = hZ = {hk : k ∈ Z}, then

f (t + h) − f (t)
σ(t) = t + h, µ(t) ≡ h, f ∆ (t) = for t ∈ T,
h
and
b
Z b h −1
X
f (t)∆t = h f (kh), where a, b ∈ T with a < b.
a a
k= h

Example 2.4. If T = Z, then

σ(t) = t + 1, µ(t) ≡ 1, f ∆ (t) = ∆f (t) = f (t + 1) − f (t) for t ∈ T,

and
Z b b−1
X
f (t)∆t = f (k), where a, b ∈ T with a < b.
a k=a

Example 2.5. Let q > 1. If T = q N0 = {q n : n ∈ N0 }, then

f (qt) − f (t)
σ(t) = qt, µ(t) = (q − 1)t, f ∆ (t) = for t ∈ T,
(q − 1)t

and
Z qn n−1
X
f (t)∆t = (q − 1) q k f (q k ), where m, n ∈ N0 with m < n.
qm k=m

3. PRIMAL AND DUAL MODELS

Throughout this paper, T stands for a time scale, we assume 0 ∈ T, we let

T ∈ T, and we use I to denote the time scales interval

I = [0, T ] ∩ T.

By Ek , we denote the space of all rd-continuous functions from I into Rk . We also

put Ek> = {f : f > ∈ Ek }. The primal time scales linear programming model is
formulated as
 Z σ(T )
 Maximize U (x) =

 f (t)x(t)∆t

 0 Z t
(P)

 subject to B(t)x(t) ≤ g(t) + K(t, s)x(s)∆s, t ∈ I

 0
and x ∈ En , x(t) ≥ 0, t ∈ I,

 Linear Programming Problems on Time Scales 195

where f ∈ En> , g ∈ Em , and B and K are rd-continuous matrix-valued functions

of size m × n. The dual time scales linear programming model is formulated as
 Z σ(T )

 Minimize V (z) =

 z(t)g(t)∆t
0

 Z σ(T )
(D)

 subject to z(t)B(t) ≥ f (t) + z(s)K(s, t)∆s, t ∈ I

 σ(t)
>

and z ∈ Em , z(t) ≥ 0, t ∈ I.


A feasible solution of (P) (or (D)) is any one that satisfies the given constraints. An
optimal solution to (P) (or (D)) is a feasible solution with the largest (or smallest)
objective function value.

4. WEAK DUALITY AND OPTIMALITY CONDITION

In this section, we state and prove the weak duality theorem and the opti-
mality condition theorem. Based on the objective functions and the constraints in
>
(P) and (D), we define the bilinear form h·, ·i1 on Em × Em by
Z σ(T )
(4.2) hz, gi1 = z(t)g(t)∆t,
0

the bilinear form h·, ·i2 on En> × En by

Z σ(T )
(4.3) hf, xi2 = f (t)x(t)∆t,
0

the linear operator A : En → Em by

Z t
(4.4) Ax(t) = B(t)x(t) − K(t, s)x(s)∆s,
0

and the linear operator A0 : Em

>
→ En> by
Z σ(T )
0
(4.5) A z(t) = z(t)B(t) − z(s)K(s, t)∆s.
σ(t)

Using these operators, we can rewrite (P) and (D) as

(
Maximize hf, xi2
(4.6)
subject to Ax ≤ g and x ≥ 0

and
(
Minimize hz, gi1
(4.7)
such that A0 z ≥ f and z ≥ 0,

respectively.
196 Rasheed Al-Salih and Martin Bohner

Lemma 4.6. We have

(4.8) hz, Axi1 = hA0 z, xi2 .

Proof. Using standard properties of the time scales integral (see [13, Section 1.4]),
we calculate
Z σ(T )
(4.2)
hz, Axi1 = z(t)Ax(t)∆t
0
Z σ(T )  Z t 
(4.4)
= z(t) B(t)x(t) − K(t, s)x(s)∆s ∆t
0 0
Z σ(T ) Z σ(T ) Z t
= z(t)B(t)x(t)∆t − z(t)K(t, s)x(s)∆s∆t
0 0 0
Z σ(T ) Z σ(T ) Z σ(T )
(∗)
= z(s)B(s)x(s)∆s − z(t)K(t, s)x(s)∆t∆s
0 0 σ(s)
" #
Z σ(T ) Z σ(T )
= z(s)B(s) − z(t)K(t, s)∆t x(s)∆s
0 σ(s)
Z σ(T )
(4.5)
= A0 z(s)x(s)∆s
0
(4.3)
= hA0 z, xi2 ,

where in (∗) we have used the time scales change of integration order formula
presented in [23, Lemma 1]. This shows (4.8) and completes the proof.
Theorem 4.7 (Weak Duality Theorem). If x and z are arbitrary feasible solutions
of (P) and (D), respectively, then U (x) ≤ V (z).

Proof. By the constraints in (4.6) and (4.7), respectively, we have

(4.9) Ax(t) ≤ g(t) and A0 z(t) ≥ f (t) for t ∈ I.

Since x(t) and z(t) are nonnegative, we obtain

(4.10) z(t)Ax(t) ≤ z(t)g(t) for t ∈ I

and

(4.11) A0 z(t)x(t) ≥ f (t)x(t) for t ∈ I.

Hence,
Z σ(T ) (4.11)
Z σ(T )
(P)
U (x) = f (t)x(t)∆t ≤ A0 z(t)x(t)∆t
0 0
Z σ(T )
(4.3) 0 (4.8) (4.2)
= hA z, xi2 = hz, Axi1 = z(t)Ax(t)∆t
0
 Linear Programming Problems on Time Scales 197

(4.10)
Z σ(T )
(D)
≤ z(t)g(t)∆t = V (z),
0

completing the proof.

Theorem 4.8 (Optimality Condition). If there exist feasible solutions x∗ and z ∗
of (P) and (D), respectively, such that U (x∗ ) = V (z ∗ ), then x∗ and z ∗ are optimal
solutions of their respective problems.

Proof. By Theorem 4.7, if x is an arbitrary feasible solution x of (P), then

U (x) ≤ V (z ∗ ) = U (x∗ ),

and thus, x∗ is an optimal solution of (P). Similarly, again by Theorem 4.7, if z is

an arbitrary feasible solution of (D), then

V (z) ≥ U (x∗ ) = V (z ∗ ),

and so, z ∗ is an optimal solution of (D).

5. STRONG DUALITY THEOREM

In this section, we consider the isolated time scale

T = {tk ∈ R : k ∈ N0 } with ti < ti+1 for all i ∈ N0 ,

where
0 = t0 < t1 < t2 < . . . < tN = T < tN +1 = σ(tN )
in both (P) and (D).
Theorem 5.9 (Strong Duality Theorem). If (P) has an optimal solution x∗ , then
(D) has an optimal solution z ∗ such that U (x∗ ) = V (z ∗ ).

Proof. Let x∗ be an optimal solution of (P), i.e., by (2.1), of


XN

Maximize U (x) = µ(tk )f (tk )x(tk )







 k=0



 subject to x(tk ) ≥ 0 for k = 0, 1, . . . , N and
B(t0 )x(t0 ) ≤g(t0 ),





B(t1 )x(t1 ) ≤g(t1 ) + µ(t0 )K(t1 , t0 )x(t0 ),
 B(t2 )x(t2 ) ≤g(t2 ) + µ(t0 )K(t2 , t0 )x(t0 ) + µ(t1 )K(t2 , t1 )x(t1 ),





 ..
.





B(t )x(t ) ≤g(t N ) + µ(t0 )K(tN , t0 )x(t0 ) + · · ·



 N N

+ µ(tN −1 )K(tN , tN −1 )x(tN −1 ),

198 Rasheed Al-Salih and Martin Bohner

i.e.,
N

 X



 Maximize U (x) = µ(tk )f (tk )x(tk )

 k=0

subject to x(tk ) ≥ 0 for k = 0, 1, . . . , N




    
x(t0 ) g(t0 )


(5.12)  x(t )   g(t ) 


  1   1 
x(t )   g(t ) 
    


 and A   2  ≤  2 ,


  ..   .. 


  .   . 

x(tN ) g(tN )


where x : I → Rn and
 
B(t0 ) 0 ··· 0
 −µ(t0 )K(t1 , t0 ) B(t1 ) ··· 0 
 
A =  −µ(t0 )K(t2 , t0 ) −µ(t1 )K(t2 , t1 ) ··· 0 .
 
 .. .. .. .. 
 . . . . 
−µ(t0 )K(tN , t0 ) −µ(t1 )K(tN , t1 ) · · · B(tN )

By the standard strong duality theorem of discrete linear programming [20, Theo-
rem 5.1 on page 58] applied to (5.12), there exists an optimal solution y ∗ of

N

 X



 Minimize W (y) = y(tk )g(tk )

 k=0

(5.13) subject to y(tk ) ≥ 0 for k = 0, 1, . . ., N




 and y(t0 ) y(t1 ) y(t2 ) · · · y(tN ) A

  
 ≥ µ(t0 )f (t0 ) µ(t1 )f (t1 ) µ(t2 )f (t2 ) · · · µ(tN )f (tN ) ,

where y > : I → Rm , and we have

(5.14) U (x∗ ) = W (y ∗ ).

Now we put

y ∗ (tk )
(5.15) z ∗ (tk ) = for k = 0, 1, . . . , N.
µ(tk )

Using (5.15) in (5.14), we get

N
X N
X
(5.16) U (x∗ ) = W (y ∗ ) = y ∗ (tk )g(tk ) = µ(tk )z ∗ (tk )g(tk ) = V (z ∗ ).
k=0 k=0
 Linear Programming Problems on Time Scales 199

Since y ∗ is feasible, it satisfies the nonnegativity constraints in (5.13), and thus

(5.15) yields
(5.17) z ∗ (t) ≥ 0 for t ∈ I,
Moreover, y ∗ also satisfies the other contraints in (5.13), and by writing them as
N + 1 inequalities and dividing the first one by µ(t0 ), the second one by µ(t1 ), . . .,
and the last one by µ(tN ), (5.15) gives



 z ∗ (t0 )B(t0 ) ≥ f (t0 ) + z ∗ (t1 )µ(t1 )K(t1 , t0 ) + · · · + z ∗ (tN )µ(tN )K(tN , t0 )

 Z σ(tN )
z ∗ (s)K(s, t0 )∆s,



 = f (t 0 ) +
σ(t )

 0

z ∗ (t1 )B(t1 ) ≥ f (t1 ) + z ∗ (t2 )µ(t2 )K(t2 , t1 ) + · · · z ∗ (tN )µ(tN )K(tN , t1 )






 Z σ(tN )
= f (t ) + z ∗ (s)K(s, t1 )∆s,



 1
σ(t1 )



z ∗ (t2 )B(t2 ) ≥ f (t2 ) + · · · z ∗ (tN )µ(tN )K(tN , t2 )

 Z σ(tN )
= f (t2 ) + z ∗ (s)K(s, t2 )∆s,





 σ(t 2 )
..



.




z ∗ (t )B(t ) ≥ f (t )



 N N N

 Z σ(tN )
z ∗ (s)K(s, tN )∆s,



 = f (tN ) +
σ(tN )

i.e.,
Z σ(T )
(5.18) z ∗ (t)B(t) ≥ f (t) + z(s)K(s, t)∆s for t ∈ I.
σ(t)

By (5.17) and (5.18), z ∗ is a feasible solution of (D). By (5.16), z ∗ is a feasible

solution of (D) satisfying V (z ∗ ) = U (x∗ ). By the optimality condition, Theorem
4.8, we conclude that z ∗ is an optimal solution of (D).

6. EXAMPLES

In this section, three examples are given in order to illustrate our duality
theorems on isolated time scales.
Example 6.10. Let T = Z and I = {0, 1, 2, 3, 4}. Then, we consider the isolated
time scales linear programming primal model
4
 Z σ(4)
 X


 Maximize U (x) = tx(t)∆t = tx(t)
0

t=0


Z t t−1
X


 subject to 6x(t) ≤ t + x(s)∆s = t + x(s), t ∈ I

 0 s=0

and x(t) ≥ 0, t ∈ I,

200 Rasheed Al-Salih and Martin Bohner

where we have used σ and the integral given in Example 2.4. Using MATLAB
command linprog or LINDO solver, we have

x∗ (0) = 0.000000, x∗ (1) = 0.166667, x∗ (2) = 0.361111,

x∗ (3) = 0.587963, x∗ (4) = 0.852623, U (x∗ ) = 6.063272.

On the other hand, the isolated time scales linear programming dual model is
 Z σ(4) X4



 Minimize V (z) = tz(t)∆t = tz(t)
0


 t=0
Z σ(T ) X4
 subject to 6z(t) ≥ t +
 z(s)∆s = t + z(s), t ∈ I
 σ(t) s=t+1



and z(t) ≥ 0, t ∈ I,


where we have used again Example 2.4. Using MATLAB command linprog or
LINDO solver, we have

z ∗ (0) = 0.382459, z ∗ (1) = 0.470679, z ∗ (2) = 0.546296,

z ∗ (3) = 0.611111, z ∗ (4) = 0.666667, V (z ∗ ) = 6.063272,

confirming U (x∗ ) = V (z ∗ ).
Example 6.11. Let T = 5Z and I = {0, 5, 10, 15, 20}. Then, we consider the isolated
time scales linear programming primal model
 Z σ(20) 4
 X
Maximize U (x) = tx(t)∆t = 25 kx(5k)





 0 k=0
 t
Z t 5 −1
X


 subject to 6x(t) ≤ t + x(s)∆s = t + 5 x(5k), t ∈ I


 0 k=0

and x(t) ≥ 0, t ∈ I,


where we have used σ and the integral given in Example 2.3. Using MATLAB
command linprog or LINDO solver, we have

x∗ (0) = 0.000000, x∗ (5) = 0.833333, x∗ (10) = 2.361111,

x∗ (15) = 5.162037, x∗ (20) = 10.297068, U (x∗ ) = 1555.748.

On the other hand, the isolated time scales linear programming dual model is
 Z σ(20) 4
 X
Minimize V (z) = tz(t)∆t = 25 kz(5k)





 0 k=0
 Z σ(T ) X 4

 subject to 6z(t) ≥ t + z(s)∆s = t + 5 z(5k), t ∈ I
 σ(t) t
k= 5 +1




and z(t) ≥ 0, t ∈ I.

 Linear Programming Problems on Time Scales 201

where we have used again Example 2.3. Using MATLAB command linprog or
LINDO solver, we have

z ∗ (0) = 27.359825, z ∗ (5) = 15.378086, z ∗ (10) = 8.842592,

z ∗ (15) = 5.277778, z ∗ (20) = 3.333333, V (z ∗ ) = 1555.748,

confirming U (x∗ ) = V (z ∗ ).
Example 6.12. Let T = 2N0 and I = {1, 2, 4}. Then, we consider the isolated time
scales linear programming primal model
 Z σ(22 )
 X2
Maximize U (x) = tx(t)∆t = 4k x(2k )






 1 k=0
Z t log2 t−1
X


 subject to 8x(t) ≤ 6t + 3 x(s)∆s = 6t + 3 2k x(2k ), t ∈ I


 1 k=0

 and x(t) ≥ 0, t ∈ I,

where we have used σ and the integral given in Example 2.5. Using MATLAB
command linprog or LINDO solver, we have

x∗ (1) = 0.750000, x∗ (2) = 1.781250,

∗
x (4) = 4.617188, U (x∗ ) = 81.75000.

On the other hand, the isolated time scales linear programming dual model is
 Z σ(22 ) X2

Minimize V (z) = 6 tz(t)∆t = 6 4k z(2k )






 1 k=0
Z σ(4) X 2

 subject to 8z(t) ≥ t + 3 z(s)∆s = t + 3 2k z(2k ), t ∈ I
σ(t)

k=1+log2 t




 and z(t) ≥ 0, t ∈ I,

where we have used again Example 2.5. Using MATLAB command linprog or
LINDO solver, we have

z ∗ (1) = 1.625000, z ∗ (2) = 1.000000,

z ∗ (2) = 0.500000, V (z ∗ ) = 81.75000,

confirming U (x∗ ) = V (z ∗ ).

7. CONCLUSIONS

In this work, duality theorems for linear programming problems on time

scales have been presented. For this novel result, a formulation of the primal and
202 Rasheed Al-Salih and Martin Bohner

the dual linear models on time scales has been established. Furthermore, the weak
duality theorem and the optimality condition theorem were given for arbitrary time
scales, while the strong duality theorem has been established for arbitrary isolated
time scales. This ensures that our formulation of primal and dual models on time
scales presented in this paper is indeed the “correct” formulation. Since all efficient
algorithms in the literature can only solve continuous-time linear programming
problems by a discrete approximation method (i.e., the optimal solution obtained
by this method is a numerical solution with an estimate of the error bound), another
contribution of this paper is to solve linear programming problems on arbitrary
isolated time scales by efficient exact methods (i.e., the optimal solution obtained
by these methods is exact).

REFERENCES

1. Ferhan M. Atici, Daniel C. Biles, and Alex Lebedinsky. An application of time scales
to economics. Math. Comput. Modelling, 43(7-8):718–726, 2006.
2. Zbigniew Bartosiewicz, Ülle Kotta, Ewa Pawluszewicz, and Malgorzata Wyrwas. Con-
trol systems on regular time scales and their differential rings. Math. Control Signals
Systems, 22(3):185–201, 2011.
3. Richard E. Bellman. Dynamic programming. Princeton Landmarks in Mathematics.
Princeton University Press, Princeton, NJ, 2010. Reprint of the 1957 edition, With a
new introduction by Stuart Dreyfus.
4. Richard E. Bellman and Stuart E. Dreyfus. Applied dynamic programming. Princeton
University Press, Princeton, N.J., 1962.
5. Martin Bohner. Calculus of variations on time scales. Dynam. Systems Appl., 13(3-
4):339–349, 2004.
6. Martin Bohner, Meng Fan, and Jimin Zhang. Periodicity of scalar dynamic equations
and applications to population models. J. Math. Anal. Appl., 330(1):1–9, 2007.
7. Martin Bohner, Rui A. C. Ferreira, and Delfim F. M. Torres. Integral inequalities and
their applications to the calculus of variations on time scales. Math. Inequal. Appl.,
13(3):511–522, 2010.
8. Martin Bohner, Gregory Gelles, and Julius Heim. Multiplier-accelerator models on
time scales. Int. J. Stat. Econ., 4(S10):1–12, 2010.
9. Martin Bohner and Svetlin G. Georgiev. Multivariable dynamic calculus on time
scales. Springer, Cham, 2016.
10. Martin Bohner and Gusein Sh. Guseinov. Double integral calculus of variations on
time scales. Comput. Math. Appl., 54(1):45–57, 2007.
11. Martin Bohner, Julius Heim, and Ailian Liu. Solow models on time scales. Cubo,
15(1):13–31, 2013.
12. Martin Bohner, Julius Heim, and Ailian Liu. Qualitative analysis of a Solow model
on time scales. J. Concr. Appl. Math., 13(3-4):183–197, 2015.
13. Martin Bohner and Allan Peterson. Dynamic equations on time scales. Birkhäuser
Boston, Inc., Boston, MA, 2001. An introduction with applications.
 Linear Programming Problems on Time Scales 203

14. Martin Bohner and Allan Peterson. Advances in dynamic equations on time scales.
Birkhäuser Boston, Inc., Boston, MA, 2003.
15. Martin Bohner and Nick Wintz. The linear quadratic regulator on time scales. Int.
J. Difference Equ., 5(2):149–174, 2010.
16. Martin Bohner and Nick Wintz. The linear quadratic tracker on time scales. Int. J.
Dyn. Syst. Differ. Equ., 3(4):423–447, 2011.
17. Martin Bohner and Nick Wintz. Controllability and observability of time-invariant
linear dynamic systems. Math. Bohem., 137(2):149–163, 2012.
18. Martin Bohner and Nick Wintz. The Kalman filter for linear systems on time scales.
J. Math. Anal. Appl., 406(2):419–436, 2013.
19. R. N. Buie and Jaromı́r Abrham. Numerical solutions to continuous linear program-
ming problems. Z. Operations Res. Ser. A-B, 17(3):A107–A117, 1973.
20. Vašek Chvátal. Linear programming. A Series of Books in the Mathematical Sciences.
W. H. Freeman and Company, New York, 1983.
21. Richard C. Grinold. Continuous programming. I. Linear objectives. J. Math. Anal.
Appl., 28:32–51, 1969.
22. Stefan Hilger. Analysis on measure chains—a unified approach to continuous and
discrete calculus. Results Math., 18(1-2):18–56, 1990.
23. Başak Karpuz. Unbounded oscillation of higher-order nonlinear delay dynamic equa-
tions of neutral type with oscillating coefficients. Electron. J. Qual. Theory Differ.
Equ., pages No. 34, 14, 2009.
24. Norman Levinson. A class of continuous linear programming problems. J. Math.
Anal. Appl., 16:73–83, 1966.
25. Ewa Pawluszewicz. Observability of nonlinear control systems on time scales. Internat.
J. Systems Sci., 43(12):2268–2274, 2012.
26. Christopher C. Tisdell and Atiya Zaidi. Basic qualitative and quantitative results
for solutions to nonlinear, dynamic equations on time scales with an application to
economic modelling. Nonlinear Anal., 68(11):3504–3524, 2008.
27. William F. Tyndall. A duality theorem for a class of continuous linear programming
problems. J. Soc. Indust. Appl. Math., 13:644–666, 1965.
28. William F. Tyndall. An extended duality theorem for continuous linear programming
problems. SIAM J. Appl. Math., 15:1294–1298, 1967.
29. Ching-Feng Wen, Yung-Yih Lur, and Hang-Chin Lai. Approximate solutions and error
bounds for a class of continuous-time linear programming problems. Optimization,
61(2):163–185, 2012.
204 Rasheed Al-Salih and Martin Bohner

Rasheed Al-Salih (Received 26.04.2017.)

Missouri S&T, (Revised 14.10.2017.)
Department of Mathematics and Statistics,
Rolla, MO 65409-0020,
USA.
E-mail: rbahhd@mst.edu

Martin Bohner
Missouri S&T,
Department of Mathematics and Statistics,
Rolla, MO 65409-0020,
USA.
E-mail: bohner@mst.edu