ASP Competition 2009

dates still subject to change!

Dear ASP, CP, SAT or SMT-researcher,

The second ASP-competition will take place in the first half of 2009 and will be run on a pool of linux machines of the DTAI-research group of the K.U.Leuven, Belgium. Details on the competition will be available on the website http://www.cs.kuleuven.be/~dtai/ASP-competition.

Just like the first ASP-competition (http://asparagus.cs.uni-potsdam.de/contest/), it will be a SAT-competition-like event. However, there will be only one problem category : "Model and Solve" ; and the competition is open for any kind of solver and language. This is a chance for each community to show the strenght of its applications and solvers and to challenge the other communities. For more details of the competition see below.

A rought time table for the competition is below (subject to change):

- Until 15/02/2000: submission and selection of benchmark problems

- From 15/02/2009 till 1/5/2009: start of "dry-run" periode. Participants install and test their solvers and programs on the K.U.Leuven pool

- 1/5/2009: Start of the tests.

- 15/9/2009: announcement of the results at the LPNMR09 conference, Potsdam, 14-18 September, (http://www.cs.uni-potsdam.de/lpnmr09/)

The first phase of the competition is the collection of benchmarks. This contest cannot be organized without some help of the research community. In the first place, we are looking for people that can contribute in creating a representative collection of benchmark problems and would be willing to provide support for these. In particular, we sollicit for the following:

- benchmark problems:  decision  or optimization problems (see below)

- a test program that can verify if a computed witness is a correct solution for the problem (and, in case of an optimisation problem, the program should be able to compute the "quality" of the answer)

- a set of instances for the problem, for use during the dry-run and for the contest itself.

- preferably a demonstration that the benchmark problem can effectively be solved.

If you could contribute such a problem, we would be most grateful. You can forward your problems to

marcd [ at ] cs.kuleuven.be

Here follow some further details. The competition is a simple "Model and Solve" competition:

- A number of benchmark problems will be made available on the website of the contest. The benchmark problems are search or optimisation problems. They will be specified in (non-ambiguous) natural language.

- Each participating team will submit a solver and a set of theories/programs modeling and solving each benchmark problem. During a "dry-run" period prior to the competition, each team will be able to install and test their solver+programs on a number of instances that will be provided for this purpose.

- In case of a search problem, a system will return either "Unsatisfiable" or "Satisfiable". In the latter case, the system should also generate a witness for the problem. The input problem instances of each benchmark problem, will be presented in the form of a set of atoms in a predefined input vocabulary. A witness will be represented in a similar format, as a set of atoms in some predefined output vocabulary.

- For an optimization problem, the input and output are similar, except that in case of "Satisfiable", a system may output a sequence of witnessses (hopefully of increasing quality). The last one in this list will be viewed as the proposed  solution.

- In the last phase of the contest, all solvers will be tested on a number of new and unknown instances for each of the benchmark problems.

- The details for the ranking still have to be worked out, but the philosophy is that the solver that solves most problems wins.

Yours sincerely

Marc Denecker