The integer approximation error in mixed-integer optimal control (Q431013)

The authors present theoretical results with applications in mixed-integer nonlinear optimal control. It is described a new proof of the fact that a trajectory exists with the strong property of integer feasibility that approximates the optimal relaxed solution arbitrarily close. In comparison with other previous methods, the authors show that a finite number of switches is sufficient. It is shown that the Sum Up Rounding strategy yields a constructive way to obtain an integer solution with a guaranteed bound of the performance loss, depending on the control discretization grid. In order to show the new theoretical properties and the effect of the rounding strategy, an example for solving a mixed-integer nonlinear optimal control is described. As well, there is presented an approximation result for the estimation of the Hausdorff distance between reachable sets. The new approximation results improve the previously known results in the sense that the approximation order is linear in the grid size.