# Optimal Control Problem for Uncertain Linear Systems with Multiple Input Delays

- Yi Jiang
^{1}, - Hongyan Yan
^{2}and - Yuanguo Zhu
^{1}Email author

**4**:5

https://doi.org/10.1186/s40467-016-0048-x

© The Author(s) 2016

**Received: **1 July 2016

**Accepted: **19 September 2016

**Published: **6 October 2016

## Abstract

Based on the concept of uncertain process, uncertain linear systems with multiple input delays are considered in this paper. An optimal problem for such systems is introduced and solved by using the equation of optimality and maximum principle. Finally, an example is given to show how to solve an uncertain optimal control model with time-delay.

## Keywords

## Introduction

The delay system was first proposed in the eighteenth century. Until the twentieth century, with the maximum principle and dynamic programming theory, delay system attracted significant interest from many researchers. In recent decades, the study of time-delay systems had a great development not only in theory but also in applications. Optimal control problems of linear time-delay systems were discussed in Basin [1] and Eller et al. [2]. Stochastic control problems with delay were studied by Bauer and Rieder [3] and Larssen and Risebro [4]. A set of infinite-dimensional differential equations could model the dynamic systems which have multiple input delays. Hamiltonian function can be used to solve the control problems with delay [5]. Moreover, the phenomenon of systems with multiple input delays are ubiquitous. It widely exists in communications. With the rapid development of information and technology, multi-input time-delay systems become a research hot spot.

In the real world, we will encounter a lot of indeterminacy; these phenomena cannot be explained by random events, such as high speed, about 50 kg. This fact prompted researchers to create a new mathematical tool. Liu [6] found an uncertainty theory through introducing an uncertain measure based on normality, self-duality, countable subadditivity, and product measure axioms. Based on an uncertain variable, uncertain process, and canonical process, Liu [7] presented uncertain differential equation. It aims to describe the evolution of dynamic uncertain systems and has widely applied in finance so far. Liu [8] assumed the stock price follows a time-homogenous uncertain differential equation and proposed the first uncertain stock model. Ji and Zhou [9] proved the existence and uniqueness of a multi-dimensional uncertain differential equation. Based on the uncertain differential equation, in 2010, Zhu [10] presented an equation of optimality for an uncertain optimal control problem by using dynamic programming. This equation plays a significant role in solving uncertain optimal control problems. It has been applied to uncertain bang-bang control problems by Xu and Zhu [11] and uncertain time-delay systems control problem by Chen and Zhu [12].

In this paper, we will introduce an uncertain linear system with multiple input delays and propose an optimal control problem subject to this system. The organization of the paper is as follows. In the “Preliminary” section, some basic concepts are reviewed. In the “Problem Statement” section, an uncertain linear system with multiple input delays is formulated, and the optimal control is derived in the problem by using the equation of optimality. In the “Example” section, we will give a numerical example of uncertain linear systems with multiple input delays to illustrate the result obtained in the previous section.

## Preliminary

Some concepts about uncertain measure, uncertain variable, and uncertain process can be found in Liu [6]. In convenience, we give some useful concepts. Let *Γ* be a nonempty set and *L* be a *σ*-algebra over *Γ*. The set function *M* defined on the *σ*-algebra *L* is called an uncertain measure if it satisfies the three axioms: *M*{*Γ*}=1; *M*{*Λ*}+*M*{*Λ*
^{
c
}}=1 for any event *Λ*∈*L*; and \(M\left \{\bigcup ^{\infty }_{i=1}\Lambda _{i}\right \}\leq \sum ^{\infty }_{i=1}M\{\Lambda _{i}\}\) for every countable sequence of events {*Λ*
_{
i
}}⊂*L*. Then, the triplet (*Γ*,*L*,*M*) is called an uncertainty space. An uncertain variable is a measurable function from an uncertainty space (*Γ*,*L*,*M*) to the set *R* of real numbers, and an uncertain vector is a measurable function from an uncertainty space to *R*
^{
n
}. The uncertainty distribution *Φ*: *R*→[0,1] of an uncertain variable *ξ* is defined by *Φ*(*x*)=*M*{*ξ*≤*x*} for any real number *x*. An uncertain process is a measurable function from *V*×(*Γ*,*L*,*M*) to the set of real numbers where *V* is an index set.

###
**Definition 1**

(Liu [8]) The uncertain variables *ξ*
_{1},*ξ*
_{2},⋯*ξ*
_{
m
} are said to be independent if \(M\{\bigcap ^{m}_{i=1}(\xi _{i}\in B_{i})\}=\min _{1\leq i\leq m}M\{\xi _{i}\in B_{i}\}\) for any Borel sets *B*
_{1},*B*
_{2},⋯*B*
_{
m
} of real numbers.

###
**Definition 2**

*ξ*is defined by

For any numbers *a* and *b*, *E*[*a*
*ξ*+*b*
*η*]=*a*
*E*[*ξ*]+*b*
*E*[*η*] if *ξ* and *η* are independent uncertain variables.

###
**Definition 3**

*C*

_{ t }is said to be a canonical process if (i)

*C*

_{0}=0 and almost all sample paths are Lipschitz continuous; (ii)

*C*

_{ t }has stationary and independent increments; (iii) every increment

*C*

_{ s+t }−

*C*

_{ s }is a normal uncertain variable with expected value 0 and variance

*t*

^{2}, denoted by \(C_{s+t}-C_{s}\thicksim N(0,t)\), whose distribution is

###
**Definition 4**

*X*

_{ t }be an uncertain process and

*C*

_{ t }be a canonical process. For any partition of closed interval [

*a*,

*b*] with

*a*=

*t*

_{1}<

*t*

_{2}<⋯<

*t*

_{ k+1}=

*b*, the mesh is written as △=max

_{1≤i≤k }∣

*t*

_{ i+1}−

*t*

_{ i }∣. Then, uncertain integral of

*X*

_{ t }with respect to

*C*

_{ t }is

###
**Definition 5**

*C*

_{ t }is a canonical process, and

*f*and

*g*are two functions. Then,

is called an uncertain differential equation. A solution is an uncertain process *X*
_{
t
} that satisfies (1) identically in *t*.

###
**Remark**

*Eq.*(1) is equivalent to the uncertain integral equation

## Problem Statement

with the initial condition *X*(*t*
_{0})=*X*
_{0}, where *t*
_{0} is the initial time. Here, *X*
_{
s
} is the state vector of *n* dimension, *u*
_{
s
} is the control vector of *m* dimension; *h*
_{
i
}>0(*i*=1,⋯,*p*) are positive time-delays; *h*= max{*h*
_{1},⋯,*h*
_{
p
}} is the maximum delay shift; and \(C_{s}=(C_{s_{1}},C_{s_{2}},\cdots,C_{s_{p}})\phantom {\dot {i}\!}\), where \(C_{s_{1}},C_{s_{2}},\cdots,C_{s_{p}}\phantom {\dot {i}\!}\) are independent canonical process. And *a*
_{0}(*s*),*a*
_{1}(*s*),*b*(*s*), and *B*
_{
i
}(*s*)(*i*=1,2,⋯,*p*) are piecewise continuous matrix functions of appropriate dimensions.

where *R*(*s*) is positive definite, *Ψ*
_{
T
},*L*(*s*) are nonnegative definite symmetric matrices, and *T*>0 is a time moment.

The optimal control problem is to find the control *u*
^{∗}(*t*)(*t*∈[*t*
_{0},*T*]) that maximizes the criterion *J*(*t*,*x*) along with the trajectory *x*
^{∗}(*t*)(*t*∈[*t*
_{0},*T*]), generated upon substituting *u*
^{∗}(*t*) into the state Eq. (2).

###
**Theorem 1**

*C*

_{ t }be a canonical process,

*u*

_{1t }be an

*n*×

*n*integrable uncertain process, and

*u*

_{2t }and

*v*

_{2t }be two n-dimensional integrable uncertain processes. Then, the n-dimensional linear uncertain differential equation

###
*Proof*

*U*

_{ t }and

*V*

_{ t }via uncertain differential equations,

*X*

_{ t }=

*U*

_{ t }

*V*

_{ t }is a solution of the uncertain differential Eq. (4). Write

*U*

_{ t }as

*d*

*U*

_{ t }=

*u*

_{1t }

*U*

_{ t }

*d*

*t*is

Taking *U*
_{0}=*I* and *V*
_{0}=*X*
_{0}, we get the solution (5). The theorem is proved. □

###
**Theorem 2**

*t*≥

*t*

_{0}is given by

*P*(

*t*) satisfies

*Q*(

*t*) is a solution of the following differential equation

*t*≥

*t*

_{0}is given by

###
*Proof*

*P*(

*T*)=2

*Ψ*

_{ T },

*Q*(

*T*)=0, and

*K*(

*T*)=0. By Eq. (15), we obtain Eqs. (6) and (7). By Eq. (16), Eq. (8) holds. Therefore,

*M*

_{ i }(

*t*) for this problem. Substituting the optimal control law Eq. (17) into the Eq. (2) gives

*t*,

*r*≥

*t*

_{0}, and

*u*

_{ t },

*T*∈

*R*

^{ n×m }and

*K*

_{1},

*K*

_{2}∈

*R*

^{ n×n }can be selected the same for any

*t*,

*r*≥

*t*

_{0}. Writing the last equality for \(x_{t+h_{i}},h_{i}>0\), we have

*T*=

*B*

_{ i }(

*t*) and using

*t*−

*h*

_{ i }instead of

*t*yields

for *t*≥*t*
_{0}+*h*
_{
i
}.

The theorem is proved. □

## Example

*a*

_{0}(

*t*)=0,

*a*

_{1}(

*t*)=1,

*B*(

*t*)=1,

*b*

_{0}(

*t*)=0,

*b*

_{1}(

*t*)=1,

*R*(

*t*)=1,

*L*(

*t*)=1,and

*Ψ*(

*T*)=1. So, we get

*M*

_{1}= exp(−0.1) and

*M*

_{2}(

*t*)=1. By Theorem 3.1, the function

*Q*(

*t*) satisfies

*Q*(

*t*)=0 for

*t*∈[0,2], and then

*K*(

*t*)=0 for

*t*∈[0,2]. So, we get that the optimal control \(u^{*}_{t}\) is

*P*(

*t*) satisfies

*S*=

*t*

_{0},

*t*

_{1},⋯

*t*

_{200}be an average partition of [0,2](i.e., 0=

*t*

_{0}<

*t*

_{1}<⋯<

*t*

_{200}=2), and

*Δ*

*t*=0.01. Thus,

*Δ*

*C*

_{ t }is a normal uncertain variable with expected value 0 and variance

*Δ*

*t*

^{2}, the distribution function is \(\Phi (x)=\left (1+\exp \left (-\frac {\pi x}{\sqrt {3}\Delta t}\right)\right)^{-1}\),

*x*∈

*R*. So, we maybe can get a sample point \(\tilde {c_{t}}\) of

*Δ*

*C*

_{ t }from \(\tilde {c_{t}}=\Phi ^{-1}(\text {rand}(0,1))\) that \(\tilde {c_{t}}=\frac {\sqrt {3}\Delta t}{-\pi }\ln \left (\frac {1}{\text {rand}(0,1)}-1\right)\). Thus,

*x*

_{ t }and

*u*

_{ t }may be given by the following iterative equations:

*j*=0,1,2,⋯,200, and \(u_{t_{j}-0.1}=0\), where

*t*

_{ j }∈[0,0.1]. The numerical solution

*P*(

*t*

_{ j }) of (23) is provided by

*j*=200,199,⋯,2,1 with

*P*(

*t*

_{200})=2.

*J*(0,

*X*

_{0})=1.067544, and the optimal controls and corresponding states are obtained in Table 1 and Figs. 1 and 2.

Numerical solutions

| 0 | 0.1 | 0.2 | 0.3 | 0.4 | 0.5 | 0.6 |
---|---|---|---|---|---|---|---|

| 1.000000 | 0.996292 | 1.008533 | 0.960251 | 0.968829 | 0.967966 | 0.967966 |

| 1.067544 | 1.067548 | 1.067554 | 1.067565 | 1.067586 | 1.067624 | 1.067692 |

| −2.033493 | −1.922702 | −2.025984 | −2.050895 | −1.952756 | −1.970269 | −1.968630 |

| 0.7 | 0.8 | 0.9 | 1.0 | 1.1 | 1.2 | 1.3 |

| 0.976390 | 0.964772 | 0.977926 | 0.944681 | 0.940976 | 0.944992 | 0.929304 |

| 1.067815 | 1.068038 | 1.068441 | 1.069075 | 1.070317 | 1.072567 | 10.7665 |

| −1.985991 | −1.962773 | −1.990284 | −1.92376 | −1.918448 | −1.930682 | −1.905861 |

| 1.4 | 1.5 | 1.6 | 1.7 | 1.8 | 1.9 | 2.0 |

| 0.906163 | 0.86307 | 0.873138 | 0.856087 | 0.784042 | 0.744824 | 0.678863 |

| 1.084087 | 1.097717 | 1.11969 | 1.164538 | 1.252615 | 1.439701 | 2.000000 |

| −1.871245 | −1.804668 | −1.862251 | −1.899024 | −1.870754 | −2.042682 | −2.586254 |

## Conclusion

In this paper, an optimal control problem for uncertain linear systems with multiple input delays was investigated. By using uncertain optimality equation and uncertain differential equation, then the optimal control of this problem was obtained. Finally, an example was used to illustrate the result of uncertain optimal control.

## Declarations

### Acknowledgements

This work is supported by the National Natural Science Foundation of China (No. 61273009, No. 61673011).

### Authors’ contributions

YJ carried out the study in the paper and drafted the first version of the manuscript. YJ and YZ designed the framework together. HY and YZ revised the first version of the manuscript. All authors read and approved the final manuscript.

### Competing interests

The authors declare that they have no competing interests.

**Open Access** This article is distributed under the terms of the Creative Commons Attribution 4.0 International License(http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.

## Authors’ Affiliations

## References

- Basin, M: Optimal control problems for time-delay systems, 380, 131–173. Springer, Berlin (2008).Google Scholar
- Eller, D, Aggarwal, JK, Banks, H: Optimal control of linear time-delay systems. IEEE Trans. Autom. Control. 14(6), 678–687 (1970).MathSciNetView ArticleGoogle Scholar
- Bauer, H, Rieder, U: Stochastic control problems with delay. Math. Methods Oper. Res. 62(3), 411–427 (2005).MathSciNetView ArticleMATHGoogle Scholar
- Larssen, B, Risebro, N: When are HJB-equation for control problems with stochastic delay equation finite dimensional?Stoch. Anal. Appl. 21, 643–671 (2003).MathSciNetView ArticleMATHGoogle Scholar
- Yong, J, Zhou, X: Stochastic controls: Hamiltonian systems and HJB equations. Springer, New York (1999).View ArticleMATHGoogle Scholar
- Liu, B: Uncertainty theory. 2nd ed. Springer-Verlag, Berlin (2007).View ArticleMATHGoogle Scholar
- Liu, B: Fuzzy process, hybrid process and uncertain process. Journal of Uncertain Systems. 2(1), 3–16 (2008).Google Scholar
- Liu, B: Some research problems in uncertainty theory. J Uncertain Syst. 3(1), 3–10 (2009).Google Scholar
- Ji, X, Zhou, J: Multi-dimensional uncertain differential equation: existence and uniqueness of solution. Fuzzy Optim. Decis. Making. 14(4), 477–491 (2015).MathSciNetView ArticleGoogle Scholar
- Zhu, Y: Uncertain optimal control with application to a portfolio selection model. Cybern. Syst. Int. J. 41(7), 535–547 (2010).View ArticleMATHGoogle Scholar
- Xu, X, Zhu, Y: Uncertain bang-bang control for continues time model. Cybern. Syst. Int. J. 43(6), 515–527 (2012).MathSciNetView ArticleMATHGoogle Scholar
- Chen, R, Zhu, Y: An optimal control model for uncertain systems with time-delay. J. Oper. Res. Soc. Jpn. 4(2013), 243–256.Google Scholar