A PAM approach to handling disruptions in real-time vehicle routing problems

During the urban distribution process, unexpected events may frequently result in disruptions to the current distribution plan, which need to be handled in real-time vehicle routing. In this paper, a knowledge-based modeling approach, PAM (disruption-handling Policies, local search Algorithms and object-oriented Modeling), is developed, which combines the scheduling knowledge of experienced schedulers with the optimization knowledge concerning models and algorithms in the field of Operations Research to obtain an effective solution in real time. Experienced schedulers can respond to different disruptions promptly with heuristic adjustment based on their experience, but their solutions may be inaccurate, inconsistent, or even infeasible. This method is limited when the problem becomes large-scale. The model-algorithm method can handle large-scale problems, but it has to predefine a specific disruption and a specific distribution state for constructing a model and algorithm, which is inflexible, time-consuming and consequently unable to promptly obtain solutions for responding to different disruptions in real time. PAM modeling approach combines the advantages and eliminates the disadvantages of the two methods aforementioned. Computational experiments show that solutions achieved by this modeling approach are practical and the speed of achieving the solutions is fast enough for responding to disruptions in real time.

► A modeling approach named PAM to handling disruptions in RVRP is proposed.
► It allows objectives and constraints to be changed in the modeling process.
► Computational experiments are designed and the results are analyzed.
► Solutions achieved are better and more practical.
► PAM's running speed is fast enough for responding to disruptions in real time.