General Method of Building a Real-Time Optimization Policy for Dynamic Vehicle Routing Problem

Abstract Currently, most of the policies for the dynamic demand vehicle routing problem are based on the traditional method for static problems as there is no general method for constructing a real-time optimization policy for the case of dynamic demand. Here, a new approach based on a combination of the rules from the static sub-problem to building real-time optimization policy is proposed. Real-time optimization policy is dividing the dynamic problem into a series of static sub-problems along the time axis and then solving the static ones. The static sub-problems’ transformation and solution rules include: Division rule, batch rule, objective rule, action rule and algorithm rule, and so on. Different combinations of these rules may constitute a variety of real-time optimization policy. According to this general method, two new policies called flexible G/G/m and flexible D/G/m were developed. The competitive analysis and the simulation results of these two policies proved that both are improvements upon the best existing policy.