Non-causal models in long term planning via set contractive optimal control methods

• Authors: E. A. Galperin;I. Galperin
• Venue: Mathematical and Computer Modelling: An International Journal
• Year: 2007

The notion of consistency for optimal plans introduced in [F.E. Kydland, E.C. Prescott, Rules rather than decisions: The inconsistency of optimal plans, J. Polit. Econ. 85 (3) (1977) 473-491] is studied in relation to dynamic programming and to multi-objective optimal control. It is demonstrated that the consistency condition for an optimal sequence of n policies for periods from 1 to n corresponding to a sequence of n economic agents' decisions that together constitute an optimal plan for an agreed-upon social objective function is equivalent to a multi-objective problem with n+1 criteria. An optimal plan is consistent if and only if the corresponding (n+1)-criteria problem is balanced, and consistent optimal plans do exist in such economic situations. Otherwise, the consistency requirement generates the balance set and a set of Pareto solutions, and both can be computed as illustrated in the inflation-unemployment example extended for two periods with three objectives to optimize. The procedure of ''consistent planning for the infinite horizon'' considered in [F.E. Kydland, E.C. Prescott, Rules rather than decisions: The inconsistency of optimal plans, J. Polit. Econ. 85 (3) (1977) 473-491] is shown to concur with Bellman's principle of optimality that may be invalid in many practical cases including the case considered in [F.E. Kydland, E.C. Prescott, Rules rather than decisions: The inconsistency of optimal plans, J. Polit. Econ. 85 (3) (1977) 473-491]. Finitely causal and non-causal models are introduced for long term planning in a multi-objective (Pareto) framework, and it is demonstrated how to deal with non-causal models using new set-contractive methods for multi-objective global optimization. The results are illustrated by examples.