Constraint-based computation in robotics and resource allocation


Several companies from very different sectors have come to us with problems having a common feature: the need to deal with nonhomogeneous constraints. Techniques developed in the fields of Numerical Analysis and Operations Research require, in general, that the constraints be homogeneous. On the contrary, Artificial Intelligence techniques process constraints in a symbolic way, thus allowing to address the above type of problems. Those techniques fall into four groups: heuristic search, logic programming, temporal reasoning, and interval analysis. Members of the research team have expertise in each of these groups of techniques.

The current project will tackle four applications involving nonhomogeneous constraints: the management and control of operations in distribution networks, scheduling, routing in a multi-vehicle setting, and mechanical design based on kinematic constraints. From the analysis of these applications, the most suitable combination of techniques to deal with each of them, as well as the required extensions to the basic techniques, will be determined. Some of the envisaged extensions are: the processing of n-ary constraints, symmetry exploitation, the introduction of duration constraints, and the dynamization of algorithms to permit adding constraints on line.

The benefits derived from the project include not only the completion of the four applications and the extension of the techniques, but also the integration of research groups with complementary backgrounds.

Coordinated among: IRI (CSIC-UPC), IIIA (CSIC), LSI (UPC) Period: 31.12.99-31.12.02