This section lists some other modeling hints that may be useful to you as you design your Dispatcher application.

In problems where the same truck makes many returns to the depot, you may want to use shorter routes. For example, if you have the same truck make 10 visits to the depot, you may end up with a route with 50-100 visits. Since local search complexity is O(n2) of the length of the route, you may want to split one physical truck into several symbolic trucks. You can constrain the last visit of truck 1 to be before first visit of truck 2, and so on. This will improve local search performance.

In problems, such as taxi dispatching and concrete delivery, where vehicles go from one location to another and are either empty or full, you may want to model these as arc routing problems. However, if both visits have time windows or other constraints, then they should be modeled as two nodes. The first solution and local search will see them as a single visit since the next variable is bound.