General fractional variational problem depending on indefinite integrals (Q306362)

Let \(a,b \in \mathbb{R}\), \(a<b\), \(\gamma>0\), \(k \in \mathbb{Z}_+\), \(k-1<\gamma<k\), \(x \in [a,b]\). If \(y\in AC([a,b])\) we define \[ \begin{aligned} {}_{a}I_{x}^{\gamma}y&={1 \over {\Gamma(\gamma)}} \int_{a}^{x} (x-t)^{\gamma-1} y(t) \,dt,\\ {}_{x}I_{b}^{\gamma} y&={1 \over {\Gamma(\gamma)}} \int_{x}^{b} (t-x)^{\gamma-1} y(t) \,dt;\\ {}_{a}D_{x}^{\gamma} y&={1 \over {\Gamma(k-\gamma)}} (\int_{a}^{\cdot}(\cdot-t)^{k-\gamma-1} y(t)\, dt)^{(k)}(x),\\ {}_{x}D_{b}^{\gamma}y&={(-1)^{k} \over {\Gamma(k-\gamma)}} (\int_{\cdot}^{b}(t-\cdot)^{k-\gamma-1} y(t) \,dt)^{(k)}(x);\end{aligned} \] if \(y \in AC^k([a,b])\) we define \[ \begin{aligned} {}_{a}^{C}D_{x}^{\gamma} y&={1 \over {\Gamma(k-\gamma)}} \int_{a}^{x}(x-t)^{k-\gamma-1}y^{(k)}(t)\,dt,\\ {}_{x}^{C}D_{b}^{\gamma} y&={{(-1)^{k}} \over {\Gamma(k-\gamma)}} \int_{x}^{b}(t-x)^{k-\gamma-1} y^{(k)}(t) \,dt.\end{aligned} \] Let \(m,n \in \mathbb{Z}_+\), \(\beta \in ]0,+\infty[^{m}\), \(\alpha \in ]0,+\infty[^{n}\), \(k-1<\alpha_{k}<k\) for \(k\in\{1,\dots,n\}\), \({\mathcal F}^{n}([a,b], \mathbb{R}^{m})=\{y \in AC^{n}([a,b]), \mathbb{R}^{m}) : {}_{a}^{C}D_{\cdot}^{\alpha}y \text{ and } {}_{a}I_{\cdot}^{\beta}y \text{ are continuous}\}\), \(f:[a,b] \times \mathbb{R}^{(n+2)m} \rightarrow \mathbb{R}\), \(F:[a,b] \times \mathbb{R}^{(n+2)m+1}\rightarrow \mathbb{R}\) continuously differentiable, \(y_{a}^{k},y_{b}^{k} \in \mathbb{R}^{m}\) for \(k \in \{0,\dots,n-1\}\), \(J:\{y \in {\mathcal F}^{n}([a,b], \mathbb{R}^{m}): y^{(k)}(a)=y_{a}^{k}, y^{(k)}(b)=y_{b}^{k} \text{ for } k \in \{0,\dots,n-1\}\} \rightarrow \mathbb{R}\), \(J(y)=\int_{a}^{b}F([y](x))\,dx\), where \(\{y\}(x)=(x,y(x),{}_{a}^{C}D_{x}^{\alpha}y,{}_{a}I_{x}^{\beta}y)\) and \([y](x)=(\{y\}(x),\int_{a}^{x} f(\{y\}(t))\,dt)\). Theorem. If \(y\) is a minimizer of \(J\), then \(y\) is a solution of the equation: \[ \begin{multlined} \partial_{1+j}F([y](x))+\sum_{i=1}^{n}{}_{x}D_{b}^{\alpha_{i}}(\partial_{1+m+i+n(j-1)} F)\circ[y] +{}_{x}I_{b}^{\beta_{j}}(\partial_{1+m(1+n)+j}F)\circ[y] \\ + \int_{x}^{b}\partial_{2+m(2+n)}F([y](t))\,dt\partial_{1+j}f(\{y\}(x))\\ + \sum_{i=1}^{n}{}_{x}D_{b}^{\alpha_{i}}(\int_{\cdot}^{b} \partial_{2+m(2+n)}F([y](t))\,dt (\partial_{1+m+i+n(j-1)}f)\circ\{y\})\\ + {}_{x}I_{b}^{\beta_{j}}(\int_{\cdot}^{b} \partial_{2+m(2+n)}F([y](t))\,dt(\partial_{1+m(1+n)+j}f)\circ\{y\})=0,\quad j\in\{1,\dots,m\}.\end{multlined} \] The authors give also necessary optimality conditions if the interval of integration is a proper subset of \([a,b]\) and study the isoperimetric problem. Moreover they give a numerical method and discuss the convergence of such method.