We want to run an E-commerce where we need to route 50 vehicles over more than 1000+ locations optimally. We have used the Google OR tools and their respective examples. But, we can see that the OR tools are capable of solving the optimum routes only for 100 rows [100 locations including Depo]. So, our concern is how to implement the same algorithms to get the optimal routes between 1000+ locations for 50 vehicles.
Asked
Active
Viewed 654 times
1 Answers
2
I would just challenge the notion of optimality. You are solving a large vrp with a distance/time matrix.
The time matrix has a significant error rate. So an optimal solution may not be really optimal operationally.
Furthermore, the distance matrix will also be subject to detour, blocked roads, slowdowns...
So the routing library gives very good solution (error < 1 percent, no guarantee) in reasonable time using more than imperfect data.
What would be the use of proven optimality ?
Laurent Perron
- 8,594
- 1
- 8
- 22
-
1As I said, it does not exist. Traffic is unreliable, trips as subject to disruption and detour. Worse even, fully optimal solutions without slack will be fragile. – Laurent Perron Jun 17 '20 at 22:29
-
thanks for your immediate response. – k_afr Jun 18 '20 at 09:52