This page is not updated anymore. We just keep it online because many researchers still ask us for our data. Good luck !

Institute of Computing - UNICAMP

Scheduling Problem under Labour Constraints -- Benchmarks

How to cite this page

Introduction

The Scheduling Problem under Labour Constraints (SPLC) is one of those problems usually found in Resource Constrained Project Scheduling (RCPS). The SPLC can be described as follows:

There is a set of orders, each of one is assigned to a single machine and is split up into several identical jobs. Each job consists of a sequence of tasks of duration 1 that must be executed one immediately after the other. The tasks of a job for a given order have a demand for labour defined by a labour requirement profile. All jobs of an order must be executed in the same machine and the order cannot be preempted (one order cannot have its execution interrupted to start the execution of another order assigned to the same machine). There are precedence relations between jobs, and there is a limit in the total number of workers available, in each period, to keep the jobs running.

The goal is to find a minimum makespan schedule - a schedule that minimizes the completion time of all jobs - that satisfies job precedence relations and that respects the labour constraints (labour usage cannot exceed the limit available in each period). SPLC is NP-hard in general.

This particular RCPS problem models a practical situation which, to our knowledge, was first studied in the project PAMIPS . SPLC and other related problems are discussed in "J. Kallrath and J. M. Wilson, 1997, Business Optimisation Using Mathematical Programming, Macmillan Press, UK. ISBN 0-333-67623-8". This reference also contains a solution of the instance Ins_10o_88j_A (L=24) (see Section Instances below).

Although a lot of work can be found in the literature of the classical Resource Constrained Project Scheduling, much less is reported specifically on the SPLC. Nevertheless, some research about SPLC is being done. This includes different approaches like Mixed Integer Programming Formulation, Constraint Net Propagation and Heuristic Methods like Tabu Search.

Our purpose here is to provide some instances of the SPLC, in order to make possible the comparison between the different strategies used in the search for good solutions of this problem. By doing so, we hope to establish a communication channel between researchers interested in the SPLC and, perhaps, to help in future design and development of new approaches.

We will do our best to maintain this home page. Participants are encouraged to send us comments and suggestions. Data sets and the format of submitted results are better explained in the next sections.

Thanks for you cooperation !

Cristina C. B. Cavalcante
Cid Carvalho de Souza




Chronology

June 4, 1997
Scheduling Problem under Labour Constraints Benchmarks WWW Home Page posted at
"http://www.ic.unicamp.br/~cris/public_html/benchmarcks/SPLC.html"
September 8, 1997
CLP results on the instance data set (Section "History").
September 1998
Cristina succesfully defended her MSc. thesis on the LCSP ... :-)
... and left IC-UNICAMP. We miss her ! :-(
February 18, 2000
The page moved to
"http://www.ic.unicamp.br/~cid/SPLC/SPLC.html"




Instances

In order to generate the instances data set, we implemented a random instance generator with the following characteristics:

INPUT:


OUTPUT:


Format of the .dat files

The .dat file, output of the random instance generator, has the following format (comments lines begin with % and are not include in the data files) :


total number of orders

% One for each job
< #jobs > < first job serial number > < last job serial number > < job duration > < labour per job >
< labour profile >
< job order number > < job number in order > < job serial number > < # predecessors > < # successors >
% One for each predecessor
< job order number > < job number in order > < job serial number >
% One for each successor
< job order number > < job number in order > < job serial number >


length (in periods) of the critical path



Instance Data Files

The instances should be tested for different numbers of available workers (L). Typical values of L are 18, 21, 24, 27 and 30.



Results (how to inform new ones)

New results can be sent by mail to in the following format:


makespan
< job serial number> < job start time >
< job serial number> < job start time >
.
.
.
Name and e-mail of who obtained the new solution
Date
Method used: (LP, CLP, heuristics, hybrid, etc)
Computation time
Computer features: (memory, CPU, etc)


Best Solutions known (September 1998)

Since September 1998, better solutions have been found. A table with the best values is presently being constructed. We are not sure we could easily retrieve the solutions corresponding to these values but, we'll do our best to get them ! Thus, the maximum we could say about the solutions reported below (except of course when optimality has been proved) is that they correspond to good solutions.

History




Questions and Comments

Please, mail to for any questions and/or comments.






References

Clique here for a list of known references (in BibTex format) on the SPLC.

(see also the book cited at the beginning of this page).




Acknowledgments

We would like to express our gratitude to the following persons who had contributed with comments, results and suggestions:




Scheduling Problem under Labour Constraints -- Benchmarks WWW Home Page maintained by Cristina C. B. Cavalcante & Cid Carvalho de Souza.

Last updated February, 2000.



This research was supported by grant 96/10270-8 from .