Dynamic routing for milk-run tours with time windows in stochastic time-dependent networks

We consider finding static yet robust recurring milk-run tours while dynamically routing the vehicle between site visits. The network arcs experience recurrent congestion, leading to stochastic and time-dependent travel times. Based on vehicle location, time of day, and current and projected network congestion states, we generate dynamic routing policies (DRP) for every pair of sites using stochastic dynamic programming (SDP). By simulating DRP we find travel time distributions for each pair of sites which is used to build the robust tour using another SDP formulation. Results are very promising when the algorithms are tested in a simulated network using historical traffic data.