An abstract topological approach to dynamic programming (Q1184843)

The author uses \textit{T. C. Koopmans}' [Econometrica 28, 287--309 (1960; Zbl 0149.38401)] recursive utility framework with a utility function \(U\) defined over the space of action streams \(A^ \infty\equiv \{_t a= (a_t, a_{t+1},\ldots): a_t\in A\}\) (\(A\) denotes the action space). Then he defines a production correspondence \(T: A\to A\) and an initial action \(\overline a_0\in A\), \(T^s\) denotes the composition of \(T\) with itself \(s\) times. The key assumption is biconvergence (upper and lower convergence) of the utility function \(U\) over \(X^\infty_{s= 1} T^s(\overline a_0)\); this property is invariant under all continuous monotonic transformations of \(U\). The paper deals with a model of dynamic programming in order to maximize the utility function subject to the feasibility constraints specified by \(T\) and \(\overline a_ 0\). In the analysis of the model, Bellman's principle of optimality is used and the true value function is calculated as the limit of successive approximations. The author also proves an asymptotic sufficient condition for biconvergence which is more easily tested and more general than the Lipschitz condition. The author believes that this approach can be very useful, especially for economic applications, because it is intuitive and easy.