Application of constraint programming technology for planning action in emergency situations

The authors developed a technology of AI planning, focused on the study of poorly formalized subject domains, knowledge about which is quantitative and qualitative. The technology provides support for the domain model open for operational modifications, allowing the inclusion / exclusion of restrictions, quality criteria, as well as setting the initial and target states using undetermined parameters. The problem of AI planning is proposed to be set and solved in the framework of an objectoriented extension of the programming technology, which places increased demands on the efficiency of processing qualitative constraints. It is proposed to present the qualitative constraints in the form of specialized matrix-like structures, and their processing should be carried out using the author's methods of non-numerical constraint satisfaction. The proposed approach allows structuring semantically closely coupled quantitative and qualitative constraints, simplifying their maintenance, as well as speeding up their automatic generation and processing. As an example of an applied task, the paper considers a simplified version of the task of planning actions for the localization of a territorial spill of petroleum products.