these days a customer asked whether it is sufficient to interrupt a planning call after checking whether two succeeding iterations do return almost same quality. Well: you risk to loose overall quality. So if you have time just spend it and stay patient.
The screenshot demonstrates the effect of the improvement phase. Within our search for the global optimum (the highest summit in the mountains) we risk to check less promising neighbourhood areas (climb downhill temporarily). And in this sample the patience was worth it.
- Planning Scenario with 200 orders and 8 sufficient vehicles. The diagrams show the change of the KPIs within 120 seconds of improvement. After 40 seconds the objective function flips from 950'000 to 850'000 (delta about 10%). So though it looks like "we are done" after 30 seconds there might be a significant impact after a period of "no change".
Bernd
#progress #charts
