Constraint reasoning techniques, developed in the context of AI research, have successfully solved many real problems. Although it is a mature field, there are many important active lines of research such as open CSPs, distributed CSPs, quantified CSPs and soft constraint problems. Another well-known AI topic is Planning. It has recently gained new strenth in the research community due to its reformulation in terms of graphs, search and constraint satisfaction. In this project we want to make progress in the open lines of research of both fields. We also want to exploit the close relation between constraint processing and planning. Our objectives are:
Constraint reasoning. Our goal is to contribute to the efficient resolution of the new paradigms (open, distributed and quantified CSPs) as well as continue our work on soft constraints, where we have already made relevant contributions, extending the results to closely related models such as clausal formulas or bayesian networks. Since the problem is in general untractable, we also want to identify tractable classes (soluble in polynomial time).
Our goal is to refine the methods based on heuristic search for planning. We also want to study and develop the combination of search and inference. We want to exploit the existence of symmetries to decrease the solving effort, which is a well-known topic in the constraint community.
We present this project as a continuation of REPLI (TIC2002-04470-C03) motivated by the good results and experience that was achieved. We have now a larger and more experienced team of researchers. The benefits of the project will be the accomplishment of the previous goals as well as the integration of different research groups with complementary expertise.