Optimal oscillations in control models: How can constant demand lead to cyclical production ? (Q1820419)

The paper deals with an infinite horizon, autonomous control problem with a one-dimensional state variable and an m-vector of controls. The evolution equation is an ODE. Introducing costs of control switches it is shown that an optimal solution (if it exists) becomes either monotonic or periodic after a sufficiently long interval of time. Another approach is to make the control continuous by considering it as a second state variable, the rate of change of which can be controlled. A cost of such control is introduced by a convex adjustment cost. Using the Hopf bifurcation theorem yields a sufficient condition for existence of stable limit cycles, expressed in terms of the model parameters. An example in production planning is given to illustrate the two approaches.