Hi there,
I am using jsprit to solve a VRP with capacity constraint, time window and skill. When I run my code several times, I do not get the same results while I fix number of iterations to 500 and I check passed data are always the same. Here is the cost value I got for three different runs :
1535.3953715421521
1525.4840781101238
1531.67723409477
here is the code :
for (int j = 0; j < 3 ; j++) {
VehicleRoutingProblemSolution solution = solve(orders, vehs);
}
What happens if you run the same experiment with many more iterations, like 5000?
even with more iteration it is not stable.
This could be for a number of reasons - I’m not completely familiar with the ruin-and-recreate (https://github.com/graphhopper/jsprit/blob/master/docs/Meta-Heuristic.md) method used, but it probably has some randomness involved to generate new solutions. were the costs closer together with the higher iterations?
well, as far as I know random aspect of jsprit is static, so I doubte if it is caused by the random aspect. And the behouviour by increasing the number of iterations is not predictable (I mean does not follow specific pattern).