Optimal Control Decomposition

Origin. Lev Pontryagin's maximum principle (1956); applied to economic planning via Kantorovich and Nemchinov; hierarchical decomposition via Mesarovic.

Mechanism. Converts a constrained optimization over time into a sequence of local decisions by introducing shadow prices (Lagrange multipliers) that encode the future value of current resources. The optimal trajectory satisfies the Hamiltonian conditions locally at each instant; no global foresight is required once the shadow prices are known. Hierarchical systems decompose the problem: the upper level sets prices, the lower level optimizes against those prices, and the iteration converges to the global optimum under convexity.

Procedure. State the objective function over time and the constraints. Compute the Hamiltonian: instantaneous objective plus shadow prices times the rates of change of state variables. Local decisions maximize the Hamiltonian at each instant. If the problem has hierarchical structure, assign pricing authority to the coordinator and operational authority to the subsystems. The coordinator adjusts prices until subsystem plans are mutually feasible.

Applies to. Resource allocation over time, capacity planning, capital budgeting, and any problem where current decisions have future consequences that must be traded off.

Limitations. Requires differentiability and often convexity; discrete or combinatorial problems do not admit the Hamiltonian formulation. Shadow prices also have meaning only in equilibrium; during adjustment they are not interpretable as values, and acting on disequilibrium prices produces instability. The method is a planning tool, not a real-time control; applying it in the absence of convergence to equilibrium is a category error that Soviet planning made repeatedly.