# 142 — Multi-Objective Vehicle Routing Problem Applied to Large Scale Post Office Deliveries

The PostVRP implementation represents each street as a chain of geometries embedded in the 2D coordinate plane. (One could imagine using osmnx to generate these maps!) A probability density $D(St)$ is then assigned to each street, where more central streets have higher densities than distant ones.
Then, a graph is constructed that defines each point weight as a function of its neighboring densities, as well as an additional cost associated with crossing the street. By segmenting this graph into $n$ parts on points of high centrality (high $D$ values) and traversing with $n$ agents, it is possible to come up with local optima for the PostVRP problem.