On the Value of Objective Function Adaptation in Online Optimisation

Abstract

We analyse a dynamic variant of the vehicle routing problem with soft time windows in which an average punctuality must be guaranteed (e.g. lateness is allowed at some customer sites). The existing objective function does not support both the aspiration for punctuality and least cost so that additional efforts are necessary to achieve an acceptable punctuality level at least possible costs. Within numerical experiments it is shown that static penalties are riot adequate in such a situation but that an adaptation of the objective function before its application to the next problem instance supports the search for high quality solutions of the problem.