The local stability of an open-loop Nash equilibrium in a finite horizon differential game (Q793650)

The control system \({\dot S}_ i(t)=u_ i(t)\), \(0\leq t\leq T<\infty\), \(S_ i(0)=S_{i0}\), \(i=1,...,m (S_ i\), \(u_ i\in {\mathbb{R}}^ 1)\) is considered. Necessary conditions for a control \(u^*(t)=(u^*_ 1(t),...,u^*_ m(t))\) to be a Nash equilibrium point with respect to the functionals \[ V^ i=\int^{T}_{0}e^{-rt}(R^ i(S(t))- (1/2)\quad a_ i(u_ i(t))^ 2)dt \] (S(t)\(=(S_ 1(t),...,S_ m(t))\) maximized by \(u_ i(t)\) are found (by means of the maximum principle), provided \(a_ i>0\), \(R^ i_ i>0\), \(R^ i_{ii}\leq 0\), \(R^ i_ j<0\), \(j\neq i\) where \(R^ i_ j=\partial R^ i/\partial S_ j, R^ i_{jk}=\partial^ 2R^ i/\partial S_ j\partial S_ k.\) They are \[ (1)\quad u^*_ i(t)=e^{rt}\lambda^ i(t)/a_ i,\quad(2)\quad \lambda^ i(t)=\int^{T}_{0}e^{-rt}\quad R^ i_ i(S(s))ds. \] Further it is assumed that \(m=2\) and \(R^ i_{jk}=const\). at the equilibrium. For a given initial deviation \(x^ 0_ i(t)=\Delta u^*_ i(t)\) of the equilibrium control the deviation \(\Delta \lambda^ i(t)\) and the new deviation \(x^ 1_ i(t)\) of the control are calculated (by (2) and (1)). Similarly the deviations \(x^ n_ i(t)\) of the control for \(n=2,3,..\). are defined. \(u^*(t)\) is called stable if \(x^ n_ i(t)\to 0\) in \(W^{2,2}\) for any initial deviation \(x^ 0_ i(t)\). Some inequalities sufficient for \(u^*(t)\) to be stable are given. The problem seems to make economically sense.