Stochastic maximum principle for a generalized Volterra control system (Q6912476)

The stochastic optimal control problem considered in this paper is given by the stochastic Volterra integral-diffenential equation (SVIDE)\N\[\NdX_t = b\left(t, X_t,\int_0^t k(t,s)X_sds, u_t, \int_0^t l(t,s)u_s ds\right) dt + \sigma\left(t, X_t, \int_0^t k(t,s)X_sds. u_t, \int_0^t l(t,s)u_s ds\right) dW_t,\N\]\Nwith \(0 \leq t \leq T, X_0 = x\), and the cost functional \N\[\NJ(u) = \left[ \int_0^T f\left(t, X_t, \int_0^t k(t,s)X_s ds, u_t, \int_0^t l(t,s)u_s ds\right) dt + g(X_T)\right]\N\] \Nto be minimized. Existence and uniqueness of the solution (using Gronwall's inequality) of the above SVIDE are proved. \N\NIntroducing the corresponding Hamiltonian function \N\[\NH(t,x,y,u,v,p,q) = b(t,x,y,u,v)p + \sigma(t,x,y,u,v)q + f(t,x,y,u,v), \N\]\Nthe maximum principle is formulated using \N\[\NH^*(t):= H\left(t, X_t^*, \int_0^t k(t,s) X_t^*ds, u_t^*, \int_0^t l(t,s) u_s^*ds, p_t, q_t\right), \N\]\Nwhere \((u_t^*)_{0\leq t \leq T}\) is the optimal control process, \((X_t^*)_{0\leq t \leq T}\) is the corresponding state process, and \((p_t,q_t)_{0 \leq t \leq T}\) satisfies a kind of general anticipated backward stochastic differential equation, see [\textit{S. Peng} and \textit{Z. Yang}, Ann. Probab. 37, No. 3, 877--902 (2009; Zbl 1186.60053)]. \N\NWhen proving existence and uniqueness of the solution of this equation, the difficulty caused by the dependence of the equation on the entire anticipated information is overcome by dividing the time interval \([0,T]\) into an increasing number of small intervals and studying the solvability on each of these intervals, a procedure inspired by [\textit{N. El Karoui} et al., Math. Finance 7, No. 1, 1--71 (1997; Zbl 0884.90035)]. \N\NThe proof is divided into several lemmas with some illuminating remarks in between. Finally the general theory is illustrated by treating the linear quadratic Volterra control problem, for which an explicit expression of the control process is obtained.