数学杂志  2023, Vol. 43 Issue (4): 307-322   PDF    
扩展功能
加入收藏夹
复制引文信息
加入引用管理器
Email Alert
RSS
本文作者相关文章
施翠云
Riemann-Liouville分数阶半线性发展型H-半变分不等式的可解性和最优控制
施翠云    
桂林理工大学南宁分校基础学科部, 广西 南宁 530001
摘要:本文研究了Hilbert空间中半线性Riemann-Liouville分数阶发展型H-半变分不等式的可解性和最优控制.首先,利用不动点理论和Clarke广义次微分性质得到半线性Riemann-Liouville分数阶发展型H-半变分不等式解的存在性.其次,在一般假设条件下证明系统的最优控制存在性.最后,给出一个例子来验证本文的主要结果.
关键词发展型H-半变分不等式    最优控制    Clarke广义次微分    Riemann-Liouville分数阶导数    
SOLVABILITY AND OPTIMAL CONTROL FOR RIEMANN-LIOUVILLE FRACTIONAL SEMILINEAR EVOLUTION HEMIVARIATIONAL INEQUALITIES
SHI Cui-yun    
School of Basic Science, Guilin University of Technology at Nanning, Nanning 530001, China
Abstract: This paper studies the solvability and optimal control for fractional semilinear evolution hemivariational inequalities with Riemann-Liouville fractional derivative in Hilbert space. First, we prove the existence of mild solutions for this problem by using a fixed point theorem and some properties of generalized Clarke subdifferential. Next, under some generally suitable hypotheses, the existence result of the optimal control to the fractional evolution hemivariational inequalities with Riemann-Liouville fractional derivative is also presented and obtained. Finally, we give an example to illustrate our main results.
Keywords: evolution hemivariational inequalities     optimal control     generalized clarke subdifferential     Riemann-Liouville fractional derivative    
1 引言

近年来, 分数阶微分方程引起了学者们的广泛关注, 可参见文献[14]. 正如Heymans和Podlubny在文献[5]中已经指出物理学中用Riemann-Liouville分数阶导数或积分来表示粘弹性初边值问题比起通常的物理解释更加合适. 正因为如此, Riemann-Liouville分数阶发展型微分包含近年来得到研究者的极大关注, 见文献[68] 及其参考文献.

H-半变分不等式首先是由希腊著名数学家与力学家Panagiotopoulos [912] 在研究非凸非光滑超势能的力学与工程问题中所提出的. 自此, 学者们发现它在粘弹性理论、经济学等领域有着广泛的应用, 研究人员对于它的单调性、拓扑性质和控制理论等方面做了大量研究并取得不错的成果, 见[1117] 及其参考文献.

另一方面, 最优控制在控制理论中扮演着非常重要的角色. 最优控制问题是为给定系统寻找一个控制法则,从而使得系统状态在某个时间点达到某个最优性准则. 为此, 许多研究者对其进行深入研究, 例如, Pantoja等人[18]研究了离散系统约束控制不等式的最优控制问题. [19, 20] 讨论了具有终端状态约束的线性二次型最优控制问题. 微分方程、变分不等式和拟变分不等式的最优控制问题是应用数学中一个不断延拓和充满活力的分支,已经得到了广泛的应用, 见[2230]及其参考文献.

近十年来, 分数阶H-半变分不等式的控制理论得到许多学者的关注, 例如, Liu和Zeng [31]研究了其能控性. 文献[32, 33]探讨了其可解性和最优控制. 但是, 所有这些成果的研究都是针对Caputo型分数阶H-半变分不等式的. Riemann-Liouville分数阶H-半变分不等式的可解性和最优控制一直是尚待解决的问题. 本文的主要目的是在适当的假设条件下研究Riemann-Liouville分数阶H-半变分不等式的可解性和最优控制.

$ H $为可分Hilbert空间. 本文主要考虑如下H-半变分不等式问题:

$ \begin{equation} \left\{\begin{array}{lll} \langle-^{L}D^{\alpha}_{t}x(t)+Ax(t)+Bu(t), v \rangle_H+F^0(t, x(t);v)\geq0, \; \forall v\in H, t\in J=[0, b], \\ I_{0^+}^{1-\alpha}x(t)|_{t=0}=x_{0}, \end{array}\right. \end{equation} $ (1.1)

其中, $ \langle\cdot, \cdot\rangle_H $表示Hilbert空间$ H $的内积, $ ^{L}D_t^{\alpha} $表示$ 0<\alpha<1 $阶Riemann-Liouville型分数阶导数. 算子$ A $是Hilbert空间$ H $中的$ C_{0}- $半群$ T(t)(t\geq 0) $的无穷小生成元. $ F^0(t, \cdot;\cdot) $表示局部Lipschitz函数$ F(t, \cdot):H\rightarrow \mathbb{R} $的广义Clarke次微分. 控制函数$ u $是取值于适当的允许控制集$ U_{ad} $中, $ B $为有界线性算子. 对于系统(1.1)下的允许状态控制对$ (x, u) $, 考虑如下指标函数

$ \begin{equation*} \mathcal {J}(x, u)=\int_J\mathcal {L}(t, x(t), u(t))dt. \end{equation*} $

本文的主要结构安排如下: 下一章将给出本文用到的一些预备知识. 第3章主要研究半线性Riemann-Liouville分数阶发展型H-半变分不等式解的存在性. 第4章在适当条件下讨论半线性Riemann-Liouville分数阶发展型H-半变分不等式的最优控制结果. 最后, 给出一个例子来验证本文的主要结果.

2 预备知识

这一章主要介绍本文用到的定义和预备知识. 对于整篇文章, 用符号$ \rightharpoonup $表示弱收敛. 设$ (H, \|\cdot\|) $是Hilbert空间, $ Y $是自反可分Banach空间. $ L_b(H, Y) $表示定义在$ H $上的取值于$ Y $中的有界线性算子全体组成的算子空间. $ C(J, H) $表示所有从$ J=[0, b] $映射到$ H $的连续函数全体, 它的元素$ x $赋以范数$ \|x\|_{C}=\sup\limits_{t\in J}\|x(t)\|_H $. 为了考虑系统(1.1)的温和解, 我们同时考虑如下空间$ C_{1-\alpha}(J, H)=\{x :t^{1-\alpha}x(t)\in C(J, H)\} $其赋予范数为$ \|x\|_{C_{1-\alpha}}=\sup\{t^{1-\alpha}\|x(t)\|_{H}:t\in J\}. $显然, $ C_{1-\alpha}(J, X) $空间为一Banach空间.

首先, 我们给出分数阶积分和导数的定义, 详见[2, 4].

定义 2.1  积分

$ I^{\alpha}_{t}f(t)=\frac{1}{\Gamma(\alpha)}\int^{t}_{0}(t-s)^{\alpha-1}f(s)ds, \; \; \; \; t>0, \; 0<\alpha<1, $

称为$ \alpha $阶Riemann-Liouville分数阶积分, 其中$ \Gamma $是gamma函数.

定义 2.2   函数$ f $$ \alpha $阶Riemann-Liouville分数阶导数定义为

$ ^LD^{\alpha}_{t}f(t)=\frac{1}{\Gamma(n-\alpha)}\frac{d^{n}}{dt^{n}}\int_{0}^{t}(t-s)^{n-\alpha-1}f(s)ds, \; n=[\alpha]+1, \; t>0, $

其中$ [\alpha] $表示$ \alpha $的整数部分.

接下来, 本文将用到如下记号: $ \mathcal {P}(X) $表示$ X $中所有非空集合组成的集族.

$ \begin{eqnarray*} &&\mathcal {P}_{f}(X)=\{S\in \mathcal {P}(X):S {\rm{ 是闭的 }}\}, \\ &&\mathcal {P}_{fb}(X)=\{S\in \mathcal {P}(X):S {\rm{ 是闭合有界的 }}\}, \\ &&\mathcal {P}_{(w)k(c)}(X)=\{S\in \mathcal {P}(X):S {\rm{ 是(弱)紧(凸)的}}\}. \end{eqnarray*} $

现在给出一些集值映射的基本定义和结果, 对于详细介绍见专著[34].

定义 2.3  设$ W, Z $为Hausdorff拓扑空间, 则

(ⅰ) 如果对于每个开子集$ D\subseteq Z $, $ F(x_0)\subseteq D $, 存在$ x_0 $的邻域$ O(x_0) $使得$ F(x)\subseteq D $对于所有的$ x\in O(x_0) $, 则称$ F $$ x_0\in X $点是上半连续的(简记为u.s.c.);

(ⅱ) 如果对于每个有界集$ V \subseteq W $都有$ F(V ) $是相对紧的, 则称$ F $是全连续的;

(ⅲ) 设$ (\Omega, \Sigma) $是可测空间, $ (E, d) $为可分度量空间. 如果对于每个闭集$ C\subseteq E $$ F^{-1}(C)=\{t\in \Omega:F(t)\cap\; C\neq\emptyset\}\in \Sigma $, 则称集值映射$ F:\Omega\rightarrow \mathcal{P}(E) $是可测的.

下面介绍局部Lipschitz函数$ h:E\rightarrow \mathbb{R} $的Clarke广义梯度的定义与相关结论.

根据文献[35], 用符号$ h^0(x; v ) $表示$ h $$ x $点沿方向$ v $的Clarke广义方向导数, 即

$ \begin{eqnarray*} h^0(x;v)=\lim\limits_{x'\rightarrow x} \sup\limits_{\lambda\rightarrow 0^+}\frac{h(x' + \lambda v)-h(x')}{\lambda}. \end{eqnarray*} $

那么$ h $在点$ x $的Clarke广义次微分, 表示为$ \partial h\subset E^* $, 由下式给出

$ \begin{array}{l} \partial h(x)=\{x^*\in X^* : h^0(x;v)\geq \langle x^*, v\rangle, \; {\text{对任意的}} \; v \in E\}. \end{array} $

以下引理是关于Clarke广义方向导数和广义梯度的基本性质, 它对证明主要结果具有重要作用.

引理 2.1  [文献[9]性质3.23, 文献[36]性质5.6.10] 设$ \Omega\subset E $为开集, $ h :\Omega\rightarrow \mathbb{R} $为局部Lipschitz函数, 则

(ⅰ) 对任意$ v \in X $, 有$ h^0(x;v) = \max\{\langle x^*, v \rangle : x^*\in \partial h(x)\} $;

(ⅱ) 对每一个$ x\in \Omega $, 梯度$ \partial h(x) $$ E^* $中的非空、凸、弱$ ^* $-紧子集,且对任意$ x^*\in h(x) $, $ \|x^*\|_{X^*}\leq \kappa $ (其中$ \kappa > 0 $$ h $$ x $附近的Lipschitz常数);

(ⅲ) 广义梯度$ \partial h $的图在$ \Omega\times E^*_{w^*} $拓扑中是闭的, 即, 如果序列$ \{x_n\}\subset \Omega $, $ \{y_n^*\}\subset E^* $使得$ y_n^*\in \partial h(x_n) $和在$ E $$ x_n\rightarrow x $, 在$ E^* $$ y_n^*\rightarrow y^* $, 则$ y^*\in \partial f(x) $ (其中$ E^*_{w^*} $表示Banach空间$ E $赋予$ w^* $-拓扑);

(ⅳ) 集值函数$ \Omega \ni x \rightarrow \partial h(x)\subseteq E $$ \Omega $映到$ X^*_{w^*} $是上半连续的.

引理 2.2   [文献[37]性质2.35] 令$ E $为可分自反的Banach空间, $ 0 < b < \infty $, 函数$ h : (0, b) \times E \rightarrow \mathbb{R} $, 使得对任意$ x \in X $, $ h(\cdot, x) $可测, 对所有$ t \in (0, b) $, $ h(t , \cdot) $局部Lipschitz连续. 则集值函数$ (0, b) \times E\ni (t , x)\rightarrow \partial h(t , x) \subset X^* $可测.

接下来, 考虑半线性Riemann-Liouville型分数阶微分包含:

$ \begin{equation} \left\{\begin{array} {l}^{L}D^{\alpha}_{t}x(t)\in Ax(t)+Bu(t)+\partial F(t, x(t)), \; t\in J=[0, b], \; 0<\alpha<1, \\ I_{0^+}^{1-\alpha}x(t)|_{t=0}=x_{0}, \end{array}\right. \end{equation} $ (2.1)

其中$ ^{L}D_t^{\alpha} $表示$ 0<\alpha<1 $阶Riemann-Liouville型分数阶导数. 算子$ A $是Hilbert空间$ H $中的$ C_{0}- $半群$ T(t)(t\geq 0) $的无穷小生成元. $ \partial F $表示局部Lipschitz函数$ F(t, \cdot):H\rightarrow \mathbb{R} $的广义Clarke次微分. 控制函数$ u $是取值于适当的允许控制集$ U_{ad} $中, $ B $为有界线性算子.

如果存在$ f\in L^p(J, H)(p>\dfrac{1}{\alpha}) $使得$ f(t)\in\partial F(t, x(t)) $

$ \begin{equation*} \left\{\begin{array} {l}^{L}D^{\alpha}_{t}x(t)=Ax(t)+Bu(t)+f(t), \; {\rm{a.e.}}\, t\in J=[0, b], \\ I_{0^+}^{1-\alpha}x(t)|_{t=0}=x_{0}, \end{array}\right. \end{equation*} $

则称$ x\in C_{1-\alpha}(J, H) $是系统(2.1) 的解. 由解的定义,有

$ \begin{equation*} \left\{\begin{array}{ll} \langle-^{L}D^{\alpha}_{t}x(t)+Ax(t)+Bu(t), v \rangle_H+\langle f(t), v\rangle_H\geq0, \; {\rm{a.e.}}\, \; t\in J=[0, b], \forall v\in H, \\ I_{0^+}^{1-\alpha}x(t)|_{t=0}=x_{0}. \end{array}\right. \end{equation*} $

由于$ f(t)\in\partial F(t, x(t)) $$ \langle f(t), v\rangle_H\leq F^0(t, x(t);v) $, 则有

$ \begin{equation*} \left\{\begin{array}{ll} \langle-^{L}D^{\alpha}_{t}x(t)+Ax(t)+Bu(t), v \rangle_H+F^0(t, x(t);v)\geq0, \; t\in J=[0, b], \forall v\in H, \\ I_{0^+}^{1-\alpha}x(t)|_{t=0}=x_{0}, \end{array}\right. \end{equation*} $

因此, 系统(1.1)与(2.1)等价.

所以, 本文接下来只需考虑微分包含系统(2.1).

由定义2.1, 2.2和文献[7, 8]的结果, 得到如下概念.

定义 2.4    对于给定的$ u\in U_{ad} $, 如果$ x\in C_{1-\alpha}(J, H), I_{0^+}^{1-\alpha}x(t)|_{t=0}=x_0 $及存在$ f\in L^p(J, H)(p>\dfrac{1}{\alpha}) $使得$ f(t)\in \partial F(t, x(t)) $ a.e. $ t\in J $

$ \begin{equation} x(t)=t^{\alpha-1}T_{\alpha}(t)x_0+\int_0^t(t-s)^{\alpha-1}T_{\alpha}(t-s)Bu(s)ds+\int_0^t(t-s)^{\alpha-1}T_{\alpha}(t-s)f(s)ds, \end{equation} $ (2.2)

则称函数$ x(\cdot) $是系统(2.1)的温和解, 其中

$ \begin{equation*} T_{\alpha}(t)=\alpha\int_0^{\infty}\theta\xi_{\alpha}(\theta)T(t^{\alpha}\theta)d\theta, \xi_{\alpha}(\theta)=\frac{1}{\alpha}\theta^{-1-\frac{1}{\alpha}}\varpi_{\alpha}\big(\theta^{-\frac{1}{\alpha}}\big)\geq0, \end{equation*} $
$ \begin{equation*} \varpi_{\alpha}(\theta)=\frac{1}{\pi}\sum\limits_{n=1}^{\infty}(-1)^{n-1}\theta^{-n\alpha-1} \frac{\Gamma(n\alpha+1)}{n!}\sin(n\pi\alpha), \, \, \theta\in(0, \infty), \end{equation*} $

$ \xi_{\alpha} $是定义在$ (0, \infty) $上的概率密度函数[38], 并且

$ \begin{equation*} \xi_{\alpha}(\theta)\geq0, \quad \theta\in(0, \infty), \quad \int_0^{\infty}\xi_{\alpha}(\theta)d\theta=1. \end{equation*} $

不难验证

$ \begin{equation} \int_0^{\infty}\theta\xi_{\alpha}(\theta)d\theta=\frac{1}{\Gamma(1+\alpha)}. \end{equation} $ (2.3)

引理 2.3  [文献[39]引理3.2–3.4] 算子$ T_{\alpha}(t) $具有如下性质:

(ⅰ) 对任意固定的$ t\geq 0, \; T_{\alpha}(t) $是线性有界算子, 即对任意$ x\in H, $

$ \begin{equation*} \|T_{\alpha}(t)x\|\leq \frac{M}{\Gamma(\alpha)}\|x\|. \end{equation*} $

(ⅱ) $ \{T_{\alpha}(t), t\geq 0\} $是强连续的.

(ⅲ) 如果$ T(t) $是紧的,则对任意$ t> 0, T_{\alpha}(t) $也是紧的.

引理 2.4   [40, 41]$ \beta>0, \; a(t) $$ [0, b] $上是非负局部可积的函数, $ b(t) $是定义在区间$ [0, b] $上非负非递减连续函数且$ b(t)\leq M $(常数), 并且函数$ y(t) $$ [0, b] $上是非负局部可积的函数, 且有

$ \begin{equation*} y(t)\leq a(t)+b(t)\int_{0}^{t}(t-s)^{\beta-1}y(s)ds, \; t\in [0, b]. \end{equation*} $

$ \begin{equation*} y(t)\leq a(t)+\int_{0}^{t}\left[\sum\limits_{n=1}^{\infty}\frac{[b(t)\Gamma(\beta)]^{n}}{\Gamma(n\beta)}(t-s)^{n\beta-1}a(s)\right]ds, \; t\in [0, b]. \end{equation*} $

另外, 如果$ a(t) $在区间$ [0, b] $上是非递减函数, 则

$ \begin{equation*} y(t)\leq a(t)E_\beta\left(b(t)\Gamma(\beta)t^{\beta}\right), \end{equation*} $

其中$ E_\beta $是定义如下的Mittag-Leffler函数

$ \begin{equation*} E_\beta(z)=\sum\limits_{k=0}^{\infty}\frac{z^{k}}{\Gamma(k\beta+1)}. \end{equation*} $

引理 2.5   [文献[42]引理3.5] 设半群$ T $$ H $中的紧$ C_0 $-半群, 则对任意$ p>\dfrac{1}{\alpha} $, 算子

$ \begin{equation*} (\Pi g)(\cdot)=\int_0^{\cdot}(\cdot-s)^{\alpha-1} T_{\alpha}(\cdot-s)g(s)ds, \quad {\rm{for }}\;g\in L^p(J, H) \end{equation*} $

$ L^p(J, H) $映射到$ C(J, H) $是紧的.

下面的结果是本文用到的重要不动点定理[43].

引理 2.6   [43]$ E $为局部凸Banach空间, 假设$ F : E \rightarrow \mathcal{P}(E) $为紧凸值的上半连续集值映射, 且存在0的一个闭邻域$ V $, 使得$ F(V) $为相对紧集. 如果集合

$ \begin{equation*} \Omega = \{x \in E: {\rm{ 对某些 }}\; \lambda > 1, \lambda x \in F(x)\} \end{equation*} $

是有界的, 那么$ F $存在不动点.

3 可解性

本章给出证明主要结果需要的可解性结果. 为此,整篇文章需要如下假设.

$ {\bf{H(A)}} $: 算子$ A $在空间$ H $中生成强连续半群$ T(t) $, $ t\geq0 $, 并且存在常数$ M\geq1 $使得$ \sup\limits_{t\in[0, \infty)}\|T(t)\|\leq M $. 对任意的$ t>0 $, $ T(t) $是紧的.

$ {\bf{H(F)}} $: $ F:J\times H\rightarrow \mathbb{R} $满足如下条件.

(ⅰ) 对所有的$ x\in H, $函数$ t\mapsto F(t, x) $是可测的;

(ⅱ) 函数$ x\mapsto F(t, x) $是局部Lipschitz的, 对a.e. $ t\in J $;

(ⅲ) 对所有的$ x\in H, $存在函数$ a\in L^p(J, \mathbb{R}^+)(p>\dfrac{1}{\alpha}) $和常数$ c>0 $使得

$ \begin{equation*} \|\partial F(t, x)\|_{H}=\sup\{\|f\|_{H}:f\in\partial F(t, x)\}\leq a(t)+ct^{1-\alpha}\|x\|_H, \end{equation*} $

$ {\bf{H(B)}}: $算子$ B\in L_b(Y, H). $

$ {\bf{H(U)}}: $集值映射$ U:J\mapsto \mathcal {P}_f(Y) $是可测的并且存在函数$ u\in L^p(J, Y)(p>\dfrac{1}{\alpha}), $使得

$ \begin{equation*} \| U(t)\|=\sup\{\| v\|:v\in U(t)\}\leq u(t), \quad {\rm{for \;a.e. }}\; t\in J. \end{equation*} $

定义允许控制集如下

$ \begin{equation*} U_{ad}=S_U^p=\{u\in L^p(J, Y):u(t)\in U(t)\; {\rm{a.e.}}\; t\in J\}, p>\frac{1}{\alpha}. \end{equation*} $

$ U_{ad}\neq\emptyset $ (见文献[34]中的命题2.1.7和引理2.3.2), 同时显然对所有的$ u\in U_{ad} $$ Bu\in L^p(J, H) $.

为了证明需要, 定义算子$ \mathcal {N}:L^q(J, H)\rightarrow \mathcal {P}_f(L^p(J, H))\, (\dfrac{1}{p}+\dfrac{1}{q}=1) $如下

$ \begin{equation*} \mathcal {N}(x)=\{w\in L^p(J, H):w(t)\in\partial F(t, x(t))\; {\rm{a.e.}}\; t\in J\}, \, \, x\in L^q(J, H). \end{equation*} $

由文献[9]中的引理5.3, 我们有.

引理 3.1    如果条件$ H(F) $成立, 则对于$ x\in L^q(J, H) $, 集合$ \mathcal {N}(x) $具有非空弱紧凸值的.

同时, 下面的引理在证明主要结果的时候也占据重要位置.

引理 3.2   [文献[17]引理11]   设$ H(F) $成立, 算子$ \mathcal{N} $满足: 若在$ L^q(J, H) $$ z_{n}\rightarrow z $, 若在$ L^p(J, H) $$ w_{n}\rightharpoonup w $$ w_{n}\in \mathcal{N}(z_{n}) $, 则$ w\in \mathcal{N}(z) $.

接下来给出系统(2.1)的一个先验估计.

引理 3.3    设条件$ H(A), H(B), H(F) $$ H(U) $成立, 则对于系统(2.1)的任意解$ x\in C_{1-\alpha}(J, H) $, 存在正数$ K $使得

$ \begin{equation} \|x\|_{C_{1-\alpha}(J, H)}\leq K. \end{equation} $ (3.1)

  设$ x $是系统(2.1)的一个温和解, 由定义2.4, 有$ f(t)\in\partial F(t, x(t)) $

$ \begin{equation*} x(t)=t^{\alpha-1}T_{\alpha}(t)x_0+\int_0^t(t-s)^{\alpha-1}T_{\alpha}(t-s)f(s)ds+\int_0^t(t-s)^{\alpha-1}T_{\alpha}(t-s)Bu(s)ds. \end{equation*} $

由已知条件, 利用引理2.3和Hölder不等式, 得到

$ \begin{eqnarray*} t^{1-\alpha}\|x(t)\|_H&\leq&\frac{M}{\Gamma(\alpha)}\|x_0\|+t^{1-\alpha}\left\|\int_{0}^{t}(t-s)^{\alpha-1}T_{\alpha}(t-s)f(s)ds\right\|\\ &&+t^{1-\alpha}\left\|\int_{0}^{t}(t-s)^{\alpha-1}T_{\alpha}(t-s)Bu(s)ds\right\|\\ &\leq&\frac{M}{\Gamma(\alpha)}\|x_0\|+\frac{Mt^{1-\alpha}}{\Gamma(\alpha)}\int_{0}^{t}(t-s)^{\alpha-1}\left[a(s)+cs^{1-\alpha}\|x(s)\|+\|Bu(s)\|\right]ds\\ &\leq&\frac{M}{\Gamma(\alpha)}\|x_0\|+\frac{Mb^{1-\alpha}}{\Gamma(\alpha)}\bigg(\frac{p-1}{p\alpha-1}\bigg)^{1-\frac{1}{p}} b^{\alpha-\frac{1}{p}}\bigg(\|a\|_{L^{p}}+\|Bu\|_{L^{p}}\bigg)\\ &&+\frac{Mcb^{\alpha-1}}{\Gamma(\alpha)}\int_{0}^{t}(t-s)^{\alpha-1}s^{1-\alpha}\|x(s)\|_Hds. \end{eqnarray*} $

$ W(t)=t^{1-\alpha}\|x(t)\|_H $, 则

$ \begin{eqnarray*} W(t)&\leq&\frac{M}{\Gamma(\alpha)}\|x_0\|+\frac{Mcb^{\alpha-1}}{\Gamma(\alpha)}\bigg(\frac{p-1}{p\alpha-1}\bigg)^{1-\frac{1}{p}}b^{\alpha-\frac{1}{p}}\bigg(\|a\|_{L^{p}}+\|Bu\|_{L^{p}}\bigg)\\ &&+\frac{Mcb^{1-\alpha}}{\Gamma(\alpha)}\int_{0}^{t}(t-s)^{\alpha-1}s^{1-\alpha}\|x(s)\|_Hds. \end{eqnarray*} $

由Gronwall不等式(见引理2.4), 得到

$ \begin{eqnarray*} W(t)&\leq& \left\{\frac{M}{\Gamma(\alpha)}\|x_0\|+\frac{Mcb^{\alpha-1}}{\Gamma(\alpha)}\bigg(\frac{p-1}{p\alpha-1}\bigg)^{1-\frac{1}{p}}b^{\alpha-\frac{1}{p}}\bigg(\|a\|_{L^{p}}+\|Bu\|_{L^{p}}\bigg)\right\} E_{\alpha}(Mct^{1-\alpha})\\ &\leq& \left\{\frac{M}{\Gamma(\alpha)}\|x_0\|+\frac{Mcb^{\alpha-1}}{\Gamma(\alpha)}\bigg(\frac{p-1}{p\alpha-1}\bigg)^{1-\frac{1}{p}}b^{\alpha-\frac{1}{p}}\bigg(\|a\|_{L^{p}}+\|Bu\|_{L^{p}}\bigg)\right\} E_{\alpha}(Mcb^{1-\alpha})\\ &:=&K. \end{eqnarray*} $

所以存在常数$ K>0 $使得$ \|x\|_{C_{1-\alpha}(J, H)}=\sup\limits_{t\in J}W(t)\leq K $. 证明完毕.

下面给出本章的主要结果.

定理 3.1   设条件$ H(A), H(B), H(F) $$ H(U) $成立. 则系统(2.1) 在$ J $中至少存在一个温和解.

  对任意的$ \varepsilon>0 $, 考虑集值映射$ \digamma:C_{1-\alpha}(J, H)\rightarrow \mathcal {P}_f(C_{1-\alpha}(J, H)) $如下

$ \begin{eqnarray} \digamma(x)&=&\bigg{\{} h\in C_{1-\alpha}(J, H):\; h(t)= t^{1-\alpha}T_{\alpha}(t)x_{0}+\int_{0}^{t}(t-s)^{\alpha-1}T_{\alpha}(t-s)f(s)ds\\& & +\int_{0}^{t}(t-s)^{\alpha-1}T_{\alpha}(t-s)Bu(s)ds, \; {\rm{ with }}\; f\in \mathcal{N}(x)\bigg{\}}, \; \; {\rm{ 对于 }}\; x\in C_{1-\alpha}(J, H), \end{eqnarray} $

其中$ u(t)\in U_{ad}. $显然要证明系统(2.1)存在温和解只需证明$ \digamma $具有不动点. 接下来证明$ \digamma $满足引理2.6的所有条件. 首先, 由$ \mathcal{N}(x) $的凸性可知$ \digamma(x) $对于每个$ x\in C_{1-\alpha}(J, H) $是凸的. 接下来将分五步进行证明.

步骤 1. $ \digamma $$ C_{1-\alpha}(J, H) $中从有界集映到有界集.

对于任意$ x\in B_{r}=\{x\in C_{1-\alpha}(J, H): \|x(t)\|_{C_{1-\alpha} (J, H)} \leq r\}, r>0. $实际上, 对任意的$ x\in B_{r}, \; \varphi\in\digamma(x) $, 存在$ f\in \mathcal{N}(x) $使得

$ \begin{eqnarray} \varphi(t)= t^{\alpha-1}T_{\alpha}(t)x_{0}+\int_{0}^{t}(t-s)^{\alpha-1}T_{\alpha}(t-s)f(s)ds +\int_{0}^{t}(t-s)^{\alpha-1}T_{\alpha}(t-s)Bu(s)ds, \; \; t\in J. \end{eqnarray} $

由已知条件$ H(A) $, $ H(F) $和Hölder不等式得到

$ \begin{eqnarray*} t^{1-\alpha}\|\varphi(t)\|&\leq& \|T_{\alpha}(t)x_{0}\|+t^{1-\alpha}\int_{0}^{t}(t-s)^{\alpha-1}\|T_{\alpha}(t-s)f(s)\|ds\nonumber\\ &&+t^{1-\alpha}\int_{0}^{t}(t-s)^{\alpha-1}\|T_{\alpha}(t-s)Bu(s)\|ds\nonumber\\ &\leq& \frac{M}{\Gamma(\alpha)}\|x_{0}\|+\frac{Mt^{1-\alpha}}{\Gamma(\alpha)}\int_{0}^{t}(t-s)^{\alpha-1}[a(s)+c\|x(s)\|_{X}]ds\\ &&+ \frac{Mt^{1-\alpha}}{\Gamma(\alpha)}\int_{0}^{t}(t-s)^{\alpha-1}\|Bu(s)\|ds\nonumber\\ &\leq& \frac{M}{\Gamma(\alpha)}\|x_{0}\|+\frac{Mt^{1-\alpha}}{\Gamma(\alpha)}\int_{0}^{t}(t-s)^{\alpha-1}[a(s)+cs^{\alpha-1}s^{1-\alpha}\|x(s)\|_{X}]ds\\ &&+ \frac{Mt^{1-\alpha}}{\Gamma(\alpha)}\int_{0}^{t}(t-s)^{\alpha-1}\|Bu(s)\|ds\nonumber\\ &\leq& \frac{M}{\Gamma(\alpha)}\|x_{0}\|+\frac{Mb^{1-\frac{1}{p}}}{\Gamma(\alpha)}\bigg(\frac{p-1}{p\alpha-1}\bigg)^{1-\frac{1}{p}}(\|a\|_{L^{p}}+\|Bu\|_{L^{p}}) +\frac{Mcb^{\alpha}}{\alpha\Gamma(2\alpha)}r. \end{eqnarray*} $

因此, $ \digamma(B_{r}) $$ C_{1-\alpha}(J, H) $中是有界的.

步骤 2. $ \digamma $$ B_{r} $上是等度连续的.

首先, 对任意的$ x\in B_{r}, \; \varphi\in\digamma(x) $, 存在$ f\in \mathcal{N}(x) $使得

$ \begin{eqnarray*} \varphi(t)= t^{1-\alpha}T_{\alpha}(t)x_{0}+\int_{0}^{t}(t-s)^{\alpha-1}T_{\alpha}(t-s)[f(s)+Bu(s)]ds, \; \; \; t\in J. \end{eqnarray*} $

其次, 对任意的$ x\in B_{r} $$ 0\leq \tau_{1}<\tau_{2}\leq b $, 有

$ \begin{eqnarray} &&\|\tau_{2}^{1-\alpha}\varphi(\tau_{2})-\tau_{1}^{1-\alpha}\varphi(\tau_{1})\|\\ &\leq&\|T_{\alpha}(\tau_{2})x_{0}-T_{\alpha}(\tau_{1})x_{0}\|\\ &&+\left\|\tau_{2}^{1-\alpha}\int_{0}^{\tau_{2}}(\tau_{2}-s)^{\alpha-1}T_{\alpha}(\tau_{2}-s)f(s)ds -\tau_{1}^{1-\alpha}\int_{0}^{\tau_{1}}(\tau_{1}-s)^{\alpha-1}T_{\alpha}(\tau_{1}-s)f(s)ds\right\|\\ &&+\left\|\tau_{2}^{1-\alpha}\int_{0}^{\tau_{2}}(\tau_{1}-s)^{\alpha-1}T_{\alpha}(\tau_{1}-s)Bu(s)ds -\tau_{1}^{1-\alpha}\int_{0}^{\tau_{1}}(\tau_{1}-s)^{\alpha-1}T_{\alpha}(\tau_{1}-s)Bu(s)ds\right\|\\ &\leq&\|T_{\alpha}(\tau_{2})x_{0}-T_{\alpha}(\tau_{1})x_{0}\|\\ &&+\left(\tau_{2}^{1-\alpha}-\tau_{1}^{1-\alpha}\right)\left\|\int_{0}^{\tau_{1}}(\tau_{1}-s)^{\alpha-1}T_{\alpha}(\tau_{1}-s)f(s)ds\right\|\\ &&+\tau_{2}^{1-\alpha}\left\|\int_{0}^{\tau_{1}}\left[(\tau_{2}-s)^{\alpha-1}-(\tau_{1}-s)^{\alpha-1}\right]T_{\alpha}(\tau_{1}-s)f(s)ds\right\|\\ &&+\tau_{2}^{1-\alpha}\left\|\int_{0}^{\tau_{1}}(\tau_{1}-s)^{\alpha-1}\Big[T_{\alpha}(\tau_{2}-s)-T_{\alpha}(\tau_{1}-s)\Big]f(s)ds\right\|\\ &&+\tau_{2}^{1-\alpha}\left\|\int_{\tau_{1}}^{\tau_{2}}(\tau_{2}-s)^{\alpha-1}T_{\alpha}(\tau_{2}-s)f(s)ds\right\|\\ &&+\left(\tau_{2}^{1-\alpha}-\tau_{1}^{1-\alpha}\right)\left\|\int_{0}^{\tau_{1}}(\tau_{1}-s)^{\alpha-1}T_{\alpha}(\tau_{1}-s)Bu(s)ds\right\|\\ &&+\tau_{2}^{1-\alpha}\left\|\int_{0}^{\tau_{1}}\left[(\tau_{2}-s)^{\alpha-1}-(\tau_{1}-s)^{\alpha-1}\right]T_{\alpha}(\tau_{1}-s)Bu(s)ds\right\|\\ &&+\tau_{2}^{1-\alpha}\left\|\int_{0}^{\tau_{1}}(\tau_{1}-s)^{\alpha-1}\left[T_{\alpha}(\tau_{2}-s)-T_{\alpha}(\tau_{1}-s)\right]Bu(s)ds\right\|\\ &&+\tau_{2}^{1-\alpha}\left\|\int_{\tau_{1}}^{\tau_{2}}(\tau_{2}-s)^{\alpha-1}T_{\alpha}(\tau_{2}-s)Bu(s)ds\right\|\\ &\leq& I_{1}+I_{2}+I_{3}+I_{4}+I_{5}+I_{6}+I_{7}+I_8+I_9. \end{eqnarray} $

易得

$ \begin{equation*} I_1=\|T_{\alpha}(\tau_{2})x_{0}-T_{\alpha}(\tau_{1})x_{0}\|\rightarrow0\, \, {\rm{ 当}}\, \, \tau_2\rightarrow \tau_1. \end{equation*} $

由已知条件和Hölder不等式, 有

$ \begin{eqnarray} I_{2}&\leq&\Big(\tau_{2}^{1-\alpha}-\tau_{1}^{1-\alpha}\Big)\|\int_{0}^{\tau_{1}}(\tau_{1}-s)^{\alpha-1}T_{\alpha}(\tau_{1}-s)f(s)ds\| \\ &\leq&\Big(\tau_{2}^{1-\alpha}-\tau_{1}^{1-\alpha}\Big)\Big[\frac{M}{\Gamma(\alpha)}\Big{(}\frac{p-1}{p\alpha-1}\Big{)}^{1-\frac{1}{p}} \|a\|_{L^{p}}\tau_{1}^{\alpha-\frac{1}{p}} +\frac{Mcr}{\Gamma(1+\alpha)}\tau_{1}^{\alpha}\Big]\rightarrow0\, \, {\rm{ 当}}\, \, \tau_2\rightarrow \tau_1. \end{eqnarray} $

类似地,

$ \begin{eqnarray} I_{3}&\leq&\frac{2M\tau_{2}^{1-\alpha}}{\Gamma(\alpha)}\bigg{(}\frac{p-1}{p\alpha-1}\bigg{)}^{1-\frac{1}{p}}(\|a\|_{L^{p}}+cr) (\tau_{2}-\tau_{1})^{\alpha-\frac{1}{p}}\rightarrow0\, \, {\rm{ 当}}\, \, \tau_2\rightarrow \tau_1. \\ I_{5}&\leq&\frac{M\tau_{2}^{1-\alpha}}{\Gamma(\alpha)}\bigg{(}\frac{p-1}{p\alpha-1}\bigg{)}^{1-\frac{1}{p}}(\|a\|_{L^{p}}+cr) (\tau_{2}-\tau_{1})^{\alpha-\frac{1}{p}}\rightarrow0\, \, {\rm{ 当}}\, \, \tau_2\rightarrow \tau_1. \\ I_{6}&\leq&\Big(\tau_{2}^{1-\alpha}-\tau_{1}^{1-\alpha}\Big)\frac{M}{\Gamma(\alpha)}\Big{(}\frac{p-1}{p\alpha-1}\Big{)}^{1-\frac{1}{p}} \|Bu\|_{L^{p}}\tau_{1}^{\alpha-\frac{1}{p}}\rightarrow0\, \, {\rm{ 当}}\, \, \tau_2\rightarrow \tau_1. \\ I_{7}&\leq&\frac{2M\tau_{2}^{1-\alpha}}{\Gamma(\alpha)}\bigg{(}\frac{p-1}{p\alpha-1}\bigg{)}^{1-\frac{1}{p}}\|Bu\|_{L^{p}} (\tau_{2}-\tau_{1})^{\alpha-\frac{1}{p}}\rightarrow0\, \, {\rm{ 当}}\, \, \tau_2\rightarrow \tau_1.\\ I_{9}&\leq&\frac{M\tau_{2}^{1-\alpha}}{\Gamma(\alpha)}\bigg{(}\frac{p-1}{p\alpha-1}\bigg{)}^{1-\frac{1}{p}}\|Bu\|_{L^{p}} (\tau_{2}-\tau_{1})^{\alpha-\frac{1}{p}}\rightarrow0\, \, {\rm{ 当}}\, \, \tau_2\rightarrow \tau_1. \end{eqnarray} $

$ \tau_1=0, 0<\tau_2\leq b $时, 显然$ I_4=I_8=0 $. 当$ t_1>0 $和充分小的$ \epsilon>0 $

$ \begin{eqnarray*} I_4&\leq&\tau_{2}^{1-\alpha}\left\|\int_{0}^{\tau_{1}-\epsilon}(\tau_{1}-s)^{\alpha-1}[T_{\alpha}(\tau_{2}-s) -T_{\alpha}(\tau_{1}-s)]f(s)ds\right\|\\ &&+\tau_{2}^{1-\alpha}\left\|\int_{\tau_1-\epsilon}^{\tau_{1}}(\tau_{1}-s)^{\alpha-1} [T_{\alpha}(\tau_{2}-s)-T_{\alpha}(\tau_{1}-s)]f(s)ds\right\| \nonumber \\ &\leq&\tau_{2}^{1-\alpha}\sup\limits_{s\in[0, \tau_{1}-\epsilon]}\|T_{\alpha}(\tau_{2}-s)-T_{\alpha}(\tau_{1}-s)\|\\ &&\times\left[\left(\frac{p-1}{p\alpha-1}\right)^{1-\frac{1}{p}}\|a\|_{L^{p}}(\tau_{1}^{\alpha-\frac{1}{p}}-\epsilon^{\alpha-\frac{1}{p}})+\frac{cr}{\alpha}(\tau_{1}^{\alpha}-\epsilon^{\alpha})\right]\\ &&+\frac{2M}{\Gamma(\alpha)}\left(\frac{p-1}{p\alpha-1}\right)^{1-\frac{1}{p}}\|a\|_{L^{p}} \epsilon^{\alpha-\frac{1}{p}}+\frac{2Mcr}{\Gamma(1+\alpha)}\epsilon^{\alpha}, \nonumber\\ I_8&\leq&\tau_{2}^{1-\alpha}\sup\limits_{s\in[0, \tau_{1}-\epsilon]}\|T_{\alpha}(\tau_{2}-s)-T_{\alpha}(\tau_{1}-s)\| \left(\frac{p-1}{p\alpha-1}\right)^{1-\frac{1}{p}}\|Bu\|_{L^{p}}(\tau_{1}^{\alpha-\frac{1}{p}}-\epsilon^{\alpha-\frac{1}{p}})\nonumber\\ && +\frac{2M}{\Gamma(\alpha)}\big{(}\frac{p-1}{p\alpha-1}\big{)}^{1-\frac{1}{p}}\|Bu\|_{L^{p}} \epsilon^{\alpha-\frac{1}{p}}. \end{eqnarray*} $

由半群$ T(t)(t>0) $的紧性和引理2.3可知$ T_{\alpha}(t)(t>0) $在一致算子拓扑下关于$ t $是连续的, 从而可知当$ \tau_{2}\rightarrow \tau_{1}, \; \epsilon\rightarrow0 $$ I_4 $$ I_8 $依赖于$ x\in B_{r} $趋向于0. 因此, $ \{t^{1-\alpha}(\digamma x)(t):x\in B_{r}\} $$ C_{1-\alpha}(J, H) $中是等度连续集.

步骤 3. $ \digamma $是全连续的.

对于固定的$ t\in J $, 我们将证明集合$ \Pi(t)=\{(\digamma x)(t):x\in B_{r}\} $$ H $中是相对紧的.

显然, $ \Pi(0)=\{x_{0}\} $是紧的, 因此只需考虑$ t>0 $的情形. 对于固定的$ 0<t\leq b $, 对任意的$ x\in B_{r}, \; \varphi\in\digamma(x) $, 存在$ f\in \mathcal{N}(x) $使得

$ \begin{eqnarray} \varphi(t)&=& t^{1-\alpha}T_{\alpha}(t)x_{0}+\int_{0}^{t}(t-s)^{\alpha-1}T_{\alpha}(t-s)[f(s)+Bu(s)]ds, \; \; \; t\in J. \end{eqnarray} $

对每个$ \epsilon\in(0, t), \; t\in(0, b], \; x\in B_{r} $和任意$ \delta>0 $, 定义

$ \begin{eqnarray*} \varphi^{\epsilon, \delta}(t)&=&t^{1-\alpha}T_{\alpha}(t)x_{0}+\alpha \int_{0}^{t-\epsilon}\int_{\delta}^{\infty} \theta(t-s)^{\alpha-1}\xi_{\alpha}(\theta)T((t-s)^{\alpha}\theta)[f(s)+Bu(s)]d\theta ds. \nonumber\\ &=&t^{1-\alpha}T_{\alpha}(t)x_{0}+\alpha T(\epsilon^{\alpha}\delta)\int_{0}^{t-\epsilon}\int_{\delta}^{\infty} \theta(t-s)^{\alpha-1}\xi_{\alpha}(\theta)T((t-s)^{\alpha}\theta-\epsilon^{\alpha}\delta)\\ &&\times[f(s)+Bu(s)]d\theta ds. \nonumber \end{eqnarray*} $

由算子$ T_{\alpha}(t)(t>0) $和半群$ T(\epsilon^{\alpha}\delta)\; (\epsilon^{\alpha}\delta> 0) $的紧性, 可知

$ \begin{equation*} \Pi_{\epsilon, \delta}(t)=\{\digamma^{\epsilon, \delta}(x)(t):x\in B_{r}\}, \end{equation*} $

$ H $中对每个$ \epsilon\in(0, t) $$ \delta>0 $是相对紧集. 另外,

$ \begin{eqnarray*} &&t^{1-\alpha}\|\varphi(t)-\varphi^{\epsilon, \delta}(x)(t)\|\nonumber\\ &=&t^{1-\alpha}\bigg\|\alpha\int_{0}^{t}\int_{0}^{\infty}\theta(t-s)^{\alpha-1}\xi_{\alpha}(\theta)T((t-s)^{\alpha}\theta)[f(s)+Bu(s)]d\theta ds\nonumber \\ &&-\alpha\int_{0}^{t-\epsilon}\int_{\delta}^{\infty} \theta(t-s)^{\alpha-1}\xi_{\alpha}(\theta)T((t-s)^{\alpha}\theta)[f(s)+Bu(s)]d\theta ds\bigg\|\nonumber \\ &\leq&b^{1-\alpha}\alpha M\left(\frac{p-1}{p\alpha-1}\right)^{1-\frac{1}{p}}\|a\|_{L^{p}} \left[b^{\alpha-\frac{1}{p}}\int_{0}^{\delta}\theta\xi_{\alpha}(\theta)d\theta +\frac{1}{\Gamma(1+\alpha)}\epsilon^{\alpha-\frac{1}{p}}\right]\\ &&+b^{1-\alpha}Mcr\left[\frac{1}{\Gamma(1+\alpha)}\epsilon^{\alpha} +b^{\alpha}\int_{0}^{\delta}\theta\xi_{\alpha}(\theta)d\theta\right] +b^{1-\alpha}\alpha M\left(\frac{p-1}{p\alpha-1}\right)^{1-\frac{1}{p}}\|Bu\|_{L^{p}}\\ &&\times \bigg{[}b^{\alpha-\frac{1}{p}}\int_{0}^{\delta}\theta\xi_{\alpha}(\theta)d\theta +\frac{b^{\frac{1}{p}}}{\Gamma(1+\alpha)}\epsilon^{\alpha-\frac{1}{p}}\bigg{]}. \nonumber \end{eqnarray*} $

由于$ \int_{0}^{\infty}\xi_{\alpha}(\theta)d\theta=1 $, 当$ \epsilon\rightarrow 0 $$ \delta\rightarrow 0 $时最后一个不等式趋于0. 因此, 存在相对紧集族完全覆盖集合$ \Pi(t)\; (t>0) $. 所以$ \Pi(t)\; (t>0) $$ H $中也是相对紧的. 由Ascoli-Arzelá 定理可知其是全连续的.

步骤 4. $ \digamma $是u.s.c. 为了达到这目的, 由文献[9]的命题3.3.12 (2)知, 只需证明$ \digamma $具有闭图像.

设在$ C_{1-\alpha}(J, H) $$ x_{n}\rightarrow x_{*} $, $ \varphi_{n}\in \digamma(x_{n}) $且在$ C_{1-\alpha}(J, H) $$ \varphi_{n}\rightarrow \varphi_{*} $. 我们将证明$ \varphi_{*}\in \digamma(x_{*}). $为此, $ \varphi_{n}\in \digamma(x_{n}) $意味着存在$ f_{n}\in \mathcal{N}(x_{n}) $使得

$ \begin{eqnarray} \varphi_{n}(t)= t^{1-\alpha}T_{\alpha}(t)x_{0}+\int_{0}^{t}(t-s)^{\alpha-1}T_{\alpha}(t-s)[f_{n}(s)+Bu(s)]ds. \end{eqnarray} $ (3.2)

由步骤1, 可知$ \{f_{n}\}_{n\geq1}\subseteq L^{p}(J, H) $是有界的. 因此可以假设, 如果有必要取子序列,

$ \begin{equation} f_{n}\rightharpoonup f_{*}, \; \; {\rm{ 对于某些 }}\; f_{*}\in L^{p}(J, H). \end{equation} $ (3.3)

由(3.2), (3.3) 和引理3.2得到

$ \begin{eqnarray} \varphi_{*}(t)= t^{1-\alpha}T_{\alpha}(t)x_{0}+\int_{0}^{t}(t-s)^{\alpha-1}T_{\alpha}(t-s)f_{*}(s)ds +\int_{0}^{t}(t-s)^{\alpha-1}T_{\alpha}(t-s)Bu(s)ds. \end{eqnarray} $ (3.4)

注意到在$ C_{1-\alpha}(J, H) $$ x_{n}\rightarrow x_{*} $$ f_{n}\in \mathcal{N}(x_{n}) $. 由引理2.3和(3.3), 得到$ f_{*}\in \mathcal{N}(x_{*}) $. 因此, 证得$ \varphi_{*}\in \digamma(x_{*}), $由此可知$ \digamma $具有闭图像.

步骤 5. 先验估计.

由步骤1–4, 得到$ \digamma $是u.s.c., 紧凸值的, 且$ \digamma(B_r) $是相对紧集. 由引理2.6, 只需要证明集合

$ \begin{equation*} \Psi=\{x\in C_{1-\alpha}(J, H):\lambda x\in \digamma(x), \; \lambda>1\} \end{equation*} $

是有界的, 就可以得到$ \digamma $存在一个不动点. 对所有的$ x\in\Omega $, 存在$ f\in \mathcal{N}(x) $使得

$ \begin{eqnarray*} x(t)= \lambda^{-1}t^{1-\alpha}T_{\alpha}(t)x_{0}+\lambda^{-1}\int_{0}^{t}(t-s)^{\alpha-1}T_{\alpha}(t-s)[f(s)+Bu(s)]ds. \end{eqnarray*} $

类似于引理3.1的证明, 可以得到

$ \begin{eqnarray*} t^{1-\alpha}\|x(t)\|_H\leq\kappa E_{\alpha}(Mcb^{1-\alpha}), \end{eqnarray*} $

其中

$ \begin{eqnarray*} \kappa= \frac{M}{\Gamma(\alpha)}\|x_0\|+\frac{Mcb^{\alpha-1}}{\Gamma(\alpha)}\bigg(\frac{p-1}{p\alpha-1}\bigg)^{1-\frac{1}{p}}b^{\alpha-\frac{1}{p}}\bigg(\|a\|_{L^{p}}+\|Bu\|_{L^{p}}\bigg). \end{eqnarray*} $

因此, 集合$ \Phi $是有界的. 由引理2.6可知, $ \digamma $至少存在一个不动点, 这表明系统(2.1)至少存在一个温和解. 证明完毕.

4 最优控制存在性

本章主要考虑如下Lagrange问题(P):

寻找如下形式的极小化指标泛函

$ \begin{equation} \mathcal{J}(x, u)=\int_0^T\mathcal {L}(t, x(t), u(t))dt, \end{equation} $ (4.1)

对于系统(1.1)的所有允许状态控制对, 即寻找一允许状态控制对$ \mathcal {J}(x^0, u^0) $使得

$ \begin{equation*} \mathcal {J}(x^0, u^0)\leq\mathcal{J}(x, u), \quad\forall(x, u)\in C_{1-\alpha}(J, H)\times U_{ad}, \end{equation*} $

其中$ x $表示系统(1.1) 相对应于控制$ u\in U_{ad} $的温和解.

为了求解问题(4.1)的存在性, 我们假设如下条件成立.

$ \bf {H(L)}: $泛函$ \mathcal{L}:J\times H\times Y\mapsto \mathbb{R}\bigcup\{\infty\} $满足:

(ⅰ) $ \mathcal{L}:J\times H\times Y\mapsto \mathbb{R}\bigcup\{\infty\} $是Borel可测的;

(ⅱ) $ \mathcal{L}(t, \cdot, \cdot) $对于a.e.$ t\in J $$ H\times Y $中是序列l.s.c的;

(ⅲ) $ \mathcal{L}(t, x, \cdot) $对每个$ x\in H $和a.e.$ t\in J $$ Y $中是凸的;

(ⅳ) 存在实常数$ a\geq0, b>0, $和非负函数$ \varphi\in L^1(J, \mathbb{R}) $使得

$ \begin{equation*} \mathcal {L}(t, x, u)\geq\varphi(t)+a\| x\|_H+b\| u\|_Y. \end{equation*} $

接下来, 给出最优控制问题(4.1)的存在性:

定理 4.1    如果定理3.1的条件和$ H(L) $成立, 则Lagrange问题(4.1) 存在一最优控制对, 即: 存在最优控制对$ (x^0, u^0)\in C_{1-\alpha}(J, H)\times U_{ad} $使得

$ \begin{equation*} \mathcal {J}(x^0, u^0)\leq\mathcal{J}(x, u), \; \forall(x, u)\in C_{1-\alpha}(J, H)\times U_{ad}. \end{equation*} $

  若$ \inf\{\mathcal{J}(x, u):(x, u)\in C_{1-\alpha}(J, H)\times U_{ad}\}=+\infty, $则易知Lagrange问题(4.1) 存在最优控制对.

不失一般性, 假设$ \inf\{\mathcal{J}(x, u):(x, u)\in C_{1-\alpha}(J, H)\times U_{ad}\}=\eta<\infty. $由条件$ H(L) $(ⅳ), 得到$ \eta>-\infty. $由下确界的定义可知, 存在一极小化可行对序列$ \{(x^m, u^m)\}\subset \mathcal {P}_{ad}\equiv\{(x, u):x $是系统(2.1)相对应于控制$ u\in U_{ad} $的温和解$ \} $, 使得当$ m\rightarrow \infty $$ \mathcal{J}(x^m, u^m)\rightarrow \eta $. 由于$ \{u^m\}\subseteq U_{ad}(m=1, 2, \cdots), \{u^m\} $是可分自反Banach空间$ L^p(J, Y) $中的有界子集, 则存在一子序列, 仍记为$ \{u^m\}, $$ u^0\in L^p(J, Y) $使得

$ \begin{equation*} u^m\rightharpoonup u^0\, \, {\rm{在}}\;L^p(J, Y). \end{equation*} $

另外, $ U_{ad} $是闭凸的, 则由Marzur引理知, $ u^0\in U_{ad}. $

设序列$ \{x^m\} $为系统(1.1) 所对应控制序列$ \{u^m\} $的温和解序列. 接下来, 将证明序列$ \{x^m\} $$ C_{1-\alpha}(J, H) $是相对紧的.

首先, 由$ \{u^m\} $的有界性和引理3.3, 可以证明存在正数$ w $使得$ \| x^m\|_{C_{1-\alpha}}\leq w, $由此知$ \| x^m\|_{C_{1-\alpha}} $是一致有界的.

其次, 证明$ \{x^m(t)\} $$ C_{1-\alpha}(J, H) $中是相对紧的.

实际上, 记$ y^m(t)=t^{1-\alpha}x^m(t) $. 对每个$ t\in J, $类似于定理3.1步骤3中的证明, 易知$ \{t^{1-\alpha}x(t):x\in B_r\} $$ C(J, H) $中是等度连续的及$ \{y(t):y(t)= t^{1-\alpha}x(t), t\in J\} $$ C(J, H) $中相对紧子集. 因此, $ \{x^m(t)\} $$ C_{1-\alpha}(J, H) $中是相对紧的. 从而, 存在$ x^0(t)\in C_{1-\alpha}(J, H) $使得

$ \begin{equation} x^m\rightarrow x^0\; \; \; {\rm{ 在}}\; C_{1-\alpha}(J, H). \end{equation} $ (4.2)

另外, 由$ H(F) $(ⅲ)可知, $ \{f^m\}\subseteq\mathcal {N}(x^m) $$ L^p(J, H) $中是有界的. 则由$ L^p(J, H) $的自反性知, 存在$ \{f^m\} $的子序列, 不失一般性, 仍记为$ \{f^m\} $, 和$ f^0\in L^p(J, H) $使得

$ \begin{equation} f^m\rightharpoonup f^0\in L^p(J, H), \end{equation} $ (4.3)

因此, 由(4.2), (4.3), 引理2.5和引理3.2, 可得

$ \begin{equation*} \int_0^{t}(t-s)^{\alpha-1}T_{\alpha}(t-s)f^m(s)ds\rightarrow \int_0^{t}(t-s)^{\alpha-1}T_{\alpha}(t-s)f^0(s)ds, \; a.e.\; t\in J. \end{equation*} $

从而, 得到

$ \begin{eqnarray} x^m(t)&=&t^{\alpha-1}T_{\alpha}(t)x_0+\int_0^t(t-s)^{\alpha-1}T_{\alpha}(t-s)[Bu^m(s)+f^m(s)]ds\\ &\rightarrow &x^0(t)=t^{\alpha-1}T_{\alpha}(t)x_0+\int_0^t(t-s)^{\alpha-1}T_{\alpha}(t-s)[Bu(s)+f^0(s)]ds \end{eqnarray} $

$ f^0\in \mathcal{N}(x^0) $, 这表明, $ x^0 $是系统(1.1) 相对应于控制$ u^0\in U_{ad} $的温和解.

注意到$ H(L) $满足Balder定理(见文献[44]的定理2.1)中的所有条件. 因此, 由Balder定理, 得到$ (x, u)\rightarrow \int_0^T\mathcal {L}(t, x(t), u(t))dt $在强拓扑$ L^1(J, X)\times L^1(J, Y) $中是序列l.s.c. 又由于$ L^p(J, H)\times L^p(J, Y)\subset L^1(J, H)\times L^1(J, Y) $, 从而$ \mathcal {J} $$ L^p(J, H)\times L^p(J, Y) $中也是序列l.s.c. 因此, $ \mathcal {J} $$ L^p(J, H)\times L^p(J, Y) $中是弱l.s.c. 又由条件$ H(L) $(ⅳ)知, $ \mathcal{J}>-\infty $, 所以$ \mathcal {J} $$ (x^0, u^0)\in C_{1-\alpha}(J, H)\times U_{ad} $中达到下确界, 这意味着

$ \eta=\lim\limits_{m\rightarrow \infty}\int_0^T\mathcal {L}(t, x^m(t), u^m(t))dt\geq\int_0^T\mathcal {L}(t, x^0(t), u^0(t))dt=\mathcal {J}(x^0, u^0)\geq\eta. $

证明完毕.

5 应用

$ X=L^{2}([0, 1], [0, \pi]) $为Hilbert空间. 考虑如下问题: 寻找控制函数$ u(t, z) $使得如下极小值泛函指标成立

$ \begin{equation} \mathcal{J}=\int_0^1\int_0^\pi |x(t, z)|^2+u(t, z)|^2dzdt. \end{equation} $ (5.1)

对于下列分数阶偏微分系统中:

$ \begin{equation} \left\{\begin{array}{llll} \langle \dfrac{\partial^{\frac{3}{5}}}{\partial t^{\frac{3}{5}}}x(t, z)+\dfrac{\partial^2}{\partial z^2} x(t, z)+Bu(t, z), v\rangle+F^0(t, z, x(t, z);v)\geq0, \\ \qquad \qquad \qquad \qquad \qquad t\in J=[0, 1], z\in[0, \pi], \\ x(t, 0)=x(t, \pi)=0, \quad t\in [0, 1], \\ x(0, z)=x_{0}(z), \qquad \quad z\in[0, \pi], \\ \end{array}\right. \end{equation} $ (5.2)

其中$ \dfrac{\partial^{\alpha}}{\partial t^{\alpha}} $表示$ \alpha=\dfrac{3}{5} $的Riemann-Liouville分数阶导数. $ x(t, z) $表示在点$ z\in [0, \pi] $$ t\in [0, 1] $的状态变量. $ F = F (t, z, \cdot) $是非光滑非凸的局部Lipschitz能量泛函. $ F^0(t, z, \cdot;\cdot) $表示的是关于第三个变量$ \nu $的广义Clarke梯度(见[35]). 一个简单例子函数$ F (\nu) =\min\{h_1(\nu), h_2(\nu)\} $满足条件H(F), 其中$ h_i: \mathbb{R}\rightarrow \mathbb{R} \; (i = 1, 2) $是凸二次函数(见[9]).

定义算子$ A:D(A)\subset X\rightarrow X $$ Ax=x_{zz} $, 其中定义域$ D(A) $给出如下

$ \begin{equation*} \{x\in X:x, x_z\; {\rm{是绝对连续的, }}\; \; x_{zz}\in X, \; x(t, 0)=x(t, \pi)=0\}. \end{equation*} $

由文献[45]可知, $ A $可以写成

$ \begin{equation*} Ax=\sum\limits_{n=1}^{\infty}n^{2}\langle x, e_n\rangle e_n, \; x\in D(A), \end{equation*} $

其中$ e_n(x)=\sqrt{\dfrac{2}{\pi}}\sin nx, \; 0\leq x\leq \pi\; (n=1, 2, \cdots) $$ X $的一组正交基, $ \langle\cdot, \cdot\rangle $表示$ L^2 $的对偶积. 显然算子$ A $是空间$ X $中如下表示的紧半群$ T(t) $, $ t>0 $的无穷小生成元$ T(t)x=\sum_{n=1}^{\infty}e^{-n^{2}t}\langle x, e_n\rangle e_n, \; x\in X. $同时, 定义无穷维空间$ U $

$ \begin{equation*} U=\left\{u:u=\sum\limits_{n=2}^{\infty}u_ne_n\quad {\rm{及 }} \sum\limits_{n=2}^{\infty}u_n^2<+\infty\right\}, \end{equation*} $

其相应的范数是$ \|u\|_U=\left(\sum\limits_{n=2}^{\infty} u_n^2\right)^{\frac{1}{2}} $. 记映射$ B\in L(U, X) $$ Bu=2u_2e_1+\sum\limits_{n=2}^{\infty}u_ne_n {\rm{for }}\; u=\sum\limits_{n=2}^{\infty}u_ne_n\in U, $由已知条件易得, $ \|B\|_{L(U, X)}\leq\sqrt{5} $.

从而, 定理4.1的条件都满足. 则问题(5.1) 可以写成形如系统(1.1)的抽象形式. 因此, 由定理4.1可知, 问题(5.1) 至少存在一状态控制函数对$ (x, u)\in X\times U $.

参考文献
[1] Glockle W G, Nonnenmacher T F. A fractional calculus approach of self-similar protein dynamics[J]. Biophys. J., 1995, 68: 46–53. DOI:10.1016/S0006-3495(95)80157-8
[2] Kilbas A A, Srivastava H M, Trujillo J J. Theory and applications of fractional differential equations[M]. Amsterdam: Elsevier Science B.V, 2006.
[3] Kumar S, Sukavanam N. Approximate controllability of fractional order semilinear systems with bounded delay[J]. J. Diff. Equ., 2012, 252(11): 6163–6174. DOI:10.1016/j.jde.2012.02.014
[4] Podlubny I. Fractional differential equations[M]. San Diego: Academic Press, 1999.
[5] Heymans N, Podlubny I. Physical interpretation of initial conditions for fractional differential equations with Riemann-Liouville fractional derivatives[J]. Rheol. Acta., 2006, 45: 765–771. DOI:10.1007/s00397-005-0043-5
[6] Bazhlekova E. Existence and uniqueness results for a fractional evolution equation in Hilbert space[J]. Fract. Calc. Appl. Anal., 2012, 15(2): 232–243. DOI:10.2478/s13540-012-0017-0
[7] Liu Z H, Li X W. Approximate controllability of fractional evolution systems with riemann-liouville fractional derivatives[J]. SIAM J. Control Optim., 2015, 53(4): 1920–1933. DOI:10.1137/120903853
[8] Zhou Y, Zhang L, Shen X H. Existence of mild solution for fractional evolution equations[J]. J. Integral Equ. Appl., 2013, 25(4): 557–586.
[9] Migórski S, Ochal A, Sofonea M. Nonlinear inclusions and hemivariational inequalities: models and analysis of contact problems[M]. New York: Springer, 2013.
[10] Motreanu D, Panagiotopoulos P D. Minimax theorem and qualitative properties of the solutions of hemivariational inequalities and applications, in: nonconvex optimization and its applications[M]. Dordrecht, Boston, London: Kluwer Academic Publishers, 1999.
[11] Panagiotopoulos P D. Hemivariational inequalities, applications in mechanics and engineering[M]. Berlin: Springer, 1993.
[12] Panagiotopoulos P D. Hemivariational inequality and Fan-variational inequality, new applications and results[J]. Atti. Sem. Mat. Fis. Univ. Modena XLIII., 1995: 159–191.
[13] Carl S, Motreanu D. Extremal solutions of quasilinear parabolic inclusions with generalized Clarke's gradient[J]. J. Diff. Equ., 2003, 191: 206–233. DOI:10.1016/S0022-0396(03)00022-6
[14] Liu Z H. A class of evolution hemivariational inequalities[J]. Nonlinear Anal.: TMA, 1999, 36: 91–100. DOI:10.1016/S0362-546X(98)00016-9
[15] Liu Z H. Anti-periodic solutions to nonlinear evolution equations[J]. J. Funct. Anal., 2010, 258: 2026–2033. DOI:10.1016/j.jfa.2009.11.018
[16] Migórski S, Ochal A. Existence of solutions for second order evolution inclusions with application to mechanical contact problems[J]. Optimization, 2006, 55(1-2): 101–120. DOI:10.1080/02331930500530187
[17] Migórski S, Ochal A. Quasi-static hemivariational inequality via vanishing acceleration approach[J]. SIAM J. Math. Anal., 2009, 41: 1415–1435. DOI:10.1137/080733231
[18] Pantoja J F A, De O, Mayne D Q. Sequential quadratic programming algrithm for discrede optimal control problem with control inequality constraints[J]. Inte. J. Control., 1991, 53: 823–836. DOI:10.1080/00207179108953650
[19] Asekov A Z. Explicit solution to a linear-quadratic optimal control problem with an arbitrary terminal[J]. J. Contemp. Mathemat. Anal., 2017, 52: 210–212.
[20] Han L, Camlibel M K, Pang J S, Heemels W P M H. A unified numerical scheme for linear-quadratic optimal control problems with joint control and state constraints[J]. Optimization Meth. Soft., 2012, 27(4-5): 761–799. DOI:10.1080/10556788.2011.593624
[21] Kisielewicz M. Differential inclutions and optimal control[M]. Dordrecht: Kluwer, 1991.
[22] Adly S, Bergounioux M, Mansour M A. Optimal control of a quasi-variational obstacle problem[J]. J. Glob. Optim., 2010, 47: 421–435. DOI:10.1007/s10898-008-9366-y
[23] Bin M J, Deng H Y, Li Y X, Zhao J. Properties of the set of admissible "state control" part for a class of fractional semilinear evolution control systems[J]. Fract. Calc. Appl. Anal., 2021, 24(4): 1275–1298. DOI:10.1515/fca-2021-0055
[24] Bin M J. Time optimal control for semilinear fractional evolution feedback control systems[J]. Optimization, 2019, 68(4): 819–832. DOI:10.1080/02331934.2018.1552956
[25] Bin M J, Liu Z H. Relaxation in nonconvex optimal control for nonlinear evolution hemivariational inequalities[J]. Nonlinear Anal. RWA., 2019, 50: 613–632. DOI:10.1016/j.nonrwa.2019.05.013
[26] Bin M J, Liu Z H. On the "bang-bang" principle for nonlinear evolution hemivariational inequalities control systems[J]. J. Math. Anal. Appl., 2019, 480(1): 123364–123364. DOI:10.1016/j.jmaa.2019.07.054
[27] Denkowski Z, Migorski S. Sensitivity of optimal solutions to control problems for systems described by hemivariational inequalities[J]. Control Cybern, 2004, 33: 211–236.
[28] Lunsford M L. Generalized variational and quasi-variational inequalities with discontinuous operators[J]. J. Math. Anal. Appl., 1997, 214: 245–263. DOI:10.1006/jmaa.1997.5625
[29] Micu S, Roventa I, Tucsnak M. Time optimal boundary controls for the heat equation[J]. J. Func. Anal., 2012, 263: 25–49. DOI:10.1016/j.jfa.2012.04.009
[30] Migórski S, Ochal A. Optimal control of parabolic hemivariational inequalities[J]. J. Global Optim., 2000, 17: 285–300. DOI:10.1023/A:1026555014562
[31] Liu Z H, Zeng B. Existence and controllability for fractional evolution inclusions of Clarke's subdifferential type[J]. Appl. Math. Comput., 2015, 257: 178–189.
[32] Jiang Y R, Huang N J. Solvability and optimal controls of fractional delay evolution inclusions with Clarke subdifferential[J]. Math. Methods Appl. Sci., 2017, 40(8): 3026–3039. DOI:10.1002/mma.4218
[33] Lu L, Liu Z H, Jiang W, Luo J L. Solvability and optimal controls for semilinear fractional evolution hemivariational inequalities[J]. Math. Methods Appl. Sci., 2016, 39(18): 5452–5464. DOI:10.1002/mma.3930
[34] Hu S, Papageorgiou N S. Handbook of multivalued analysis: volume Ⅰ theory[M]. Dordrecht Boston, London: Kluwer Academic Publishers, 1997.
[35] Clarke F H. Optimization and nonsmooth analysis[M]. Philadelphia: SIAM, 1990.
[36] Denkowski Z, Migóski S, Papageorgiou N S. An introduction to nonlinear analysis: theorey[M]. New York: Kluwer Academic/Plenum Publishers, 2003.
[37] Kulig A. Nonlinear evolution incutions and hemivariational inequalities for nonsmooth problem in contact mechanics[D]. Poland: Jagiellonian University, 2010.
[38] Mainardi F, Paradisi P, Gorenflo R. Probability distributions generated by fractional diffusion equations[J]. Physics, 2007. DOI:10.48550/arXiv.0704.0320
[39] Zhou Y, Jiao F. Existence of mild solutions for fractional neutral evolution equations[J]. Comput. Math. Appl., 2010, 59: 1063–1077.
[40] Dixon J, McKee S. Weakly singular discrete gronwall inequalities[J]. ZAMM·Z. Angew. Math. Mech., 1986, 68(11): 535–544.
[41] Henry D. Geometric theory of semilinear parabolic equations[M]. New York: Springer-Verlag, 1981.
[42] Wang J R, Zhou Y, Wei W. Optimal feedback control for semilinear fractional evolution equations in Banach spaces[J]. Syst. Contr. Letter, 2012, 61: 472–476. DOI:10.1016/j.sysconle.2011.12.009
[43] Ma T W. Topological degrees for set-valued compact vector fields in locally convex spaces[J]. Dissertationess Math., 1972, 92: 1–43.
[44] Balder E J. Necessary and sufficient condition for L1-strong-weak lower semicontinuity of integral functionals[J]. Nonlinear Anal., 1987, 11: 1399–1404. DOI:10.1016/0362-546X(87)90092-7
[45] Lightbourne J H, Rankin S M. A partial functional differential equation of Sobolev type[J]. J. Math. Anal. Appl., 1983, 93: 328–337. DOI:10.1016/0022-247X(83)90178-6