1
0
Fork 0
mirror of https://github.com/graphhopper/jsprit.git synced 2020-01-24 07:45:05 +01:00

IT tests with Jsprit algorithm

This commit is contained in:
oblonski 2015-03-12 10:13:10 +01:00
parent e8a9363f83
commit 4fefa79d43

View file

@ -23,8 +23,9 @@ To make it even more obvious, we added a job cloud surrounding
This is definitely not the solution, one would intuitively expect when looking at the problem sharply. This is definitely not the solution, one would intuitively expect when looking at the problem sharply.
Therefore, we put much effort into analysing this and designing a new algorithm (which is still based on the Therefore, we put much effort into analysing this and implementing a new algorithm (which is still based on the
schrimpf principles). We implemented two new ruin strategies: worst and cluster ruin, and we added principles of [Schrimpf et al. (2000)](http://www.sciencedirect.com/science/article/pii/S0021999199964136) and additionally inspired by [Pisinger and Ropke (2007)](http://www.sciencedirect.com/science/article/pii/S0305054805003023)).
We implemented two new ruin strategies: worst and cluster ruin, and we added
some thoroughly designed randomness to these and the existing strategies. Moreover, the new algorithm makes use of regretInsertion by default. some thoroughly designed randomness to these and the existing strategies. Moreover, the new algorithm makes use of regretInsertion by default.
On all benchmarking instances, our new algorithm performs better (or at least as good On all benchmarking instances, our new algorithm performs better (or at least as good
as) than the previous algorithms in terms of solution quality. However, it comes with higher computational costs. as) than the previous algorithms in terms of solution quality. However, it comes with higher computational costs.