Ramification of Volterra-type rough paths (Q2685136)

The authors study Stochastic Volterra equations of the form \[ u(t) = u_0 + \sum_{i=0}^d \int_0^t k(t,r) f_i(u_r) d q^i_r, \quad u_0\in\mathbb{R}^e \] where \(f_i\) are sufficiently regular vector fields on \(\mathbb{R}^e\), \(q\) is an \(\alpha\)-Hölder continuous signal and \(k\) is a suitable kernel, allowed to be singular in the diagonal \(t=r\). They develop a pathwise existence and solution theory, which includes signals of arbitrarily low regularity \(\alpha>0\), taking inspiration from Rough Paths and the previous contributions by \textit{F. A. Harang} and \textit{S. Tindel} [Stochastic Processes Appl. 142, 34--78 (2021; Zbl 1481.60204)] and \textit{F. A. Harang} et al. [Stoch. Dyn. 23, No. 1, Article ID 2350002, 50 p. (2023; Zbl 07674256)] which only covered \(\alpha>1/4\). The main idea is that, instead of the iterated integrals of a path, one now keeps track of iterated integral convolutions with the Volterra kernel. This leads to a general framework of \textit{Branched Rough Paths of Volterra type}, or simply \textit{Volterra Paths}, on which an operation corresponding to an integral convolution, called \textit{convolution product} and denoted by \(\star\), is well defined and satisfies some relevant algebraic properties. Once this is done, one can introduce a notion of controlled path à la Gubinelli and solve the Volterra equation by a fixed point argument.