I'm now working on somehow an inverse problem of an ODE:
Suppose we have a ODE on $\mathbb{R}^{n}$: $\dot{x} = f(x)$, denote the solution to the ODE starting at $a$ as $x_{f,a}$(t).
Now there is a bounded, simply connected open subset $\Omega\subset \mathbb{R}^{n}$, define for $a,b\in\Omega$ the space:
$$M(a,b) = \left\{f\in C^{\infty,1}_{L}(\Omega): x_{f,a}(T) = b\text{ and } x_{f,a}(t)\in\Omega,\forall t\in[0,T]\right\}$$
Where $C^{\infty,1}_{L}$ stands for Lipschitz continuous and smooth vector-valued functions satisfying
$$\sup \left\vert \frac{\partial^{k}_{x_{i}}f(x)-\partial^{k}_{y_{i}}f(y)}{x-y}\right\vert\leq L,\forall k\geq0$$
Apparently $M(a,b)$ is not a Hilbert space (not even a linear space), so many functional analysis tools fail for analysis on $M(a,b)$.
My question is, for any $f_1,f_2 \in M(a,b)$, is there any bound of $\Vert f_1-f_2\Vert_{W^{s,p}(\Omega)}$? Or is there any related research?
----Update 2024/09/09
I've found a article which may help: Vrabie, Ioan I. “Compactness of the Solution Operator for a Linear Evolution Equation with Distributed Measures.”, probably i've made the question complicated.
Thx.