Modeling, simulating and real execution of distributed constraints in NetLogo (DisCSP-Netlogo)

(joint work with Horia Popa, West University of Timisoara)

Distributed Constraint programming is a programming approach used to describe and solve large classes of problems such as searching, combinatorial and planning problems. This type of distributed modeling appears naturally in many problems for which the information is distributed among many agents. The use of agent-based simulation models in NetLogo for research is growing rapidly in a number of science and engineering domains, including in computer science, for example for analyzing the performances of asynchronous search techniques based on distributed constraints. In this talk we present an open-source and complete solution in NetLogo that allows modeling, simulation and evaluation of the distributed constraints using a single computer or clusters of computers, simulation or real execution. Our tool allows the use of various search techniques and also the evaluation and analysis of the performance of the asynchronous search techniques in two manners: simulation mode and real execution. We also explain our methodology for running the NetLogo models in a cluster computing environment (simulation mode) or on a network of computers (simulation real execution). This tool is aimed to allow the evaluation of distributed algorithms in conditions as similar as possible to the real situations. Some problems modeled with distributed constraints are exemplified:

* the randomly generated problem that has a structure of scale-free network (the constraint graph has a structure of scale-free network) .

* the multi-robot exploration problem.

* the randomly generated (binary) CSPs.

* the protein folding problem.