Introducing User-Derived Information in the Optimization of Highly Constrained Truck Loading

Truck loading problems from practice tend to involve a lot of constraints. In automatically solving these problems, a rule set has to be compiled that governs the generation of valid solutions. This rule set is either defined manually, or distilled automatically by analyzing solutions of human planners. Van Rijn et al. describe an approach for automatically solving these problems using a manually defined set, determined in consultation with industry. In a follow-up paper, we reported on an addition to this approach that introduced user-derived information into the mutation operator. This thesis extends the existing approach further by introducing this user-derived information into the function that transforms the used indirect representation into a solution that can be evaluated. Furthermore, a method is proposed to correct the penalty function used to evaluate candidate solutions. With these extensions we are able to generate solutions that accommodate more boxes into a container, as well as improve on quality, in the sense that the automatically generated solutions conform more to how a human planner solves these problems.