TOWARDS AN EFFICIENT NUMERICAL PROCEDURE FOR MIXED-INTEGER OPTIMAL-CONTROL

An efficient numerical procedure for solving mixed-integer optimal con trol (MIOPCON) problems is developed in this paper, which involves a d ecomposition strategy through a series of optimal control and mixed-in teger linear programming (MILP) subproblems. The optimal control probl em is defined by fully implicit differential-algebraic equations, whic h are substituted by discrete time implicit equations resulting from t he integration of the system equations by an implicit Runge-Kutta meth od. The advantage of this approach is that the dual information necess ary for the MILP master problem can be obtained directly from the adjo int variables of the optimal control primal problem. As a result the M IOPCON problem is solved only in the reduced space, enabling efficient application of the algorithm to problems described by large-scale dif ferential algebraic equations.