Linear differential games (analytic theory on the basis of alternating integration) (Q914571)

The paper continues the analytical theory of solving linear differential games of pursuit worked out by the authors in Sov. Math., Dokl. 30, 216- 218 (1984); translation from Dokl. Akad. Nauk SSSR 277, 1063-1066 (1984; Zbl 0589.90101) and Math. USSR, Sb. 59, 129-154 (1988); translation from Mat. Sb., Nov. Ser. 131(173), No.2, 131-158 (1986; Zbl 0626.90108). A closed-loop strategy u(t,z) bringing a system \(\dot z=Cz-u+v\), \(t\geq 0\), \(z(0)=z_ 0\), \(u\in P\), \(v\in Q\) (P and Q are convex compacta) onto a convex terminal set M no later than at time \(T(z_ 0)\) is constructed. Here T(z) is the minimum of \(\tau\geq 0\) such that \(e^{\tau C}\in W_{\tau}\) or (equivalently) (*) \(F(z,\tau)=0\); \(W_{\tau}\) is the alternated integral [see the first author, Mat. Sb., Nov. Ser. 112(154), 307-330 (1980; Zbl 0445.90118)], and (**) \(F(z,\tau)=- \min_{\psi}c(\psi)\), \(| \psi | =1\), where c(\(\psi\)) is the support functional for the set \(W_{\tau}-e^{\tau C}z\). Certain smoothness conditions for the boundaries of \(W_{\tau}\) and P are accepted. The strategy u(t,z) being substituted into the system equation ensures keeping (*) with \({\dot \tau}\leq 1\) along any motion. For the local analysis two cases are of importance: (i) \(\psi =\psi (z,\tau)\) providing min in (**) is not orthogonal to \(P_{\tau}=e^{\tau C}P\), and (ii) \(\psi\) is orthogonal to \(P_{\tau}\). For (i) an explicit expression for \({\dot \tau}\) leads to the condition: u(t,z) maximizes \((e^{\tau C}u,\psi)\). For (ii) \({\dot \tau}\) does not depend on u, and u(t,z) is found by differentiation of the orthogonality condition.