ilog cp optimizer vs OR-tools in Routing solver

207 views
Skip to first unread message

Henry Lee

unread,
Mar 11, 2024, 1:39:48 AM3/11/24
to or-tools-discuss
I have developed code to solve the TSP problem using IBM ILOG CP Optimizer and OR-Tools Routing Solver, for the single salesman problem with 2D coordinates of 50 or 100 cities, aiming to minimize the total distance traveled. However, you found that the solution obtained using Routing Solver's guided local search or similar methods surpasses the optimal solution obtained by running IBM ILOG for 10 minutes. What could be the reason for OR-Tools producing faster and better results compared to IBM ILOG CP Optimizer? Is it possible for there to be differences in both time and optimal solution values like this?
what do you think about this situation?
ilog 와 ortools 비교 사진.png

Laurent Perron

unread,
Mar 11, 2024, 2:20:19 AM3/11/24
to or-tools-discuss
They are very different solvers and were written by the same persons (in part).

CPO is a complete solver. The routing library uses local search, and is tuned for routing models. This it is expected to be much faster on tsp/vrp/PDP problems. 

Long answer:

CPO is a complete solver, but does not contain an optimized routing library.

Before CPO, there was ILOG Dispatcher which was the equivalent of the OR-Tools routing library, using local search implemented top of the of ILOG Solver.

Actually, most of the dispatcher team actually works at Google, and rewrote the CP solver, the local search, and the routing library.

Conclusion, CPO is a great solver, but not for TSP/VRP as dispatcher was killed about the same time as the launch of CPO.

--
You received this message because you are subscribed to the Google Groups "or-tools-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to or-tools-discu...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/or-tools-discuss/98dc9deb-5334-40fc-a16a-04158fd3f3e3n%40googlegroups.com.

Laurent Perron

unread,
Mar 11, 2024, 2:28:51 AM3/11/24
to or-tools-discuss
About the optimal solution. Your model is most likely wrong. Test it against known instances.

Henry Lee

unread,
Mar 12, 2024, 2:17:23 AM3/12/24
to or-tools-discuss
Thank you so much. If i consider 'time window' in TSP, Can i develop OR-tools' TSP solver code? I don't have idea considering time in TSP.

2024년 3월 11일 월요일 오후 3시 20분 19초 UTC+9에 laurent...@gmail.com님이 작성:

Laurent Perron

unread,
Mar 12, 2024, 2:56:59 AM3/12/24
to or-tools-discuss
Please read the guides, examples before askig basic questions.

In particular there is a tsptw example, and a vrptw, and a pdptw one...

Henry Lee

unread,
Mar 12, 2024, 3:00:49 AM3/12/24
to or-tools-discuss
Thank you. Sorry for my ignorance.

2024년 3월 12일 화요일 오후 3시 56분 59초 UTC+9에 Laurent Perron님이 작성:

Laurent Perron

unread,
Mar 12, 2024, 3:11:56 AM3/12/24
to or-tools-discuss
Ignorance is fine. We are all ignorant. Your first question is completely justified.

The problem is a general trend where users prefer spamming a mailing list with more than 3000 users rather than read the readme file that points to the available resources.

Reply all
Reply to author
Forward
0 new messages