Authors:
Branko Miloradović
;
Baran Çürüklü
;
Mikael Ekström
and
Alessandro Vittorio Papadopoulos
Affiliation:
Division of Intelligent Future Technologies, Mälardalen University, Högskoleplan 1, 721 23 Västerås and Sweden
Keyword(s):
Multi-Agent Mission Planning, Colored Traveling Salesperson (CTSP), Genetic Algorithms.
Related
Ontology
Subjects/Areas/Topics:
Applications
;
Methodologies and Technologies
;
Operational Research
;
OR in Transportation
;
Pattern Recognition
;
Resource Allocation
;
Routing
;
Scheduling
;
Software Engineering
Abstract:
In recent years, multi-agent systems have been widely used in different missions, ranging from underwater to airborne. A mission typically involves a large number of agents and tasks, making it very hard for the human operator to create a good plan. A search for an optimal plan may take too long, and it is hard to make a time estimate of when the planner will finish. A Genetic algorithm based planner is proposed in order to overcome this issue. The contribution of this paper is threefold. First, an Integer Linear Programming (ILP) formulation of a novel Extensive Colored Traveling Salesperson Problem (ECTSP) is given. Second, a new objective function suitable for multi-agent mission planning problems is proposed. Finally, a reparation algorithm to allow usage of common variation operators for ECTSP has been developed.