How can we prove that Quintiq's optimization technology is number one?

We've broken some of the toughest optimization puzzles in the world. Here are our latest wins:
Featured content

Flexible Job Shop Scheduling Problem

What's the problem
The Flexible Job Shop Scheduling Problem (FJSSP) is an extension of the classical Job Shop Scheduling Problem (JSSP). The JSSP is based on the assumption that only one machine is able to run a particular operation. The FJSSP, as its name suggests, assumes a more flexible setup. It reflects a production environment where it’s possible to run an operation on more than one machine.

The FJSSP was devised by Paolo Brandimarte in 1993. Brandimarte’s original FJSSP included several instances, which varied according to the flexibility of the machines. In the years since then, scientists have published multiple variations, some with additional constraints.

Though the specifications vary, the essence of the problem remains the same – you must assign all operations to machines in sequence (respecting the sequence of operations in each job) such that the total completion time is minimized.

Each instance of the FJSSP has the following characteristics:
  • The number of jobs and number of machines are specified
  • Each job consists of a fixed sequence of operations
  • Certain operations can only be performed on certain machines
  • The processing time of an operation can differ depending on the machine it runs on
  • A machine can only run one operation at a time

Quintiq's world records

Quintiq’s optimization team investigated the published instances and results of FJSSP. The team ran all 313 instances where both the specifications and results had been published. Of those 313 instances, the Quintiq optimizer has found new best-known solutions for 110 instances, ahead of any academic or industrial research group. Of these 110 best-known solutions, 90 have been proven to be optimal.

Here are some of our latest and most impressive wins:
Instance Former World Record1 Date Quintiq's World Record2 Date
Dauzere 14a 2162 Jan 2014 2161 May 2014
Hurink-vdata vdata-la25 752 Jan 2014 751 May 2014
Hurink-vdata vdata-la24 774 Nov 2013 773 Feb 2014
Dauzere 15a 2162 Jan 2014 2161 Jan 2014
Hurink-vdata vdata-abz7 493 Nov 2013 492 Jan 2014
Hurink-vdata vdata-car1 5006 Jan 2014 5005 Jan 2014
Hurink-vdata vdata-la26 1054 Feb 2000 1052 Sep 2013
Hurink-rdata rdata-la03 478 Feb 2000 477 Aug 2013
Dauzere 03a 2229 Oct 2008 2228 Mar 2013
1 All records displayed show improvements on the upper bounds. View all results, showing upper and lower bound improvements, here.
2 All results shown are optimal results – meaning, no more improvement is possible.