Route planning in a multi-agent system (MAS) is still a complex task, especially if there is need for a continuous, decentralized planner of the routes for physical agents in a dynamic environment. It is a planner of this kind that is required in the application the article considers: the transportation of parts at a modular manufacturing line. Such a planner has to meet several difficult requirements, regarding the physical, time-constrained dynamic environment, live-locks and deadlocks, delayed agents, and last needs to minimize the travel time and the total distance traveled. The article proposes an approach using a delegate multi-agent system (D-MAS) in order to meet these requirements. The approach was verified using virtual reality so as to provide a better understanding of the planner’s issues. Several coordination rules were proposed and implemented. As a further verification, the proof-of-concept solution was compared to a non-reservation planner. It was shown, that as the number of agents increases, the approach, including the reservations, outperformed its competitor. Various recommendations for the implementation of the planner were formulated. It was concluded that the performance of the planner is sufficient for its future use. The main objective of article was proof-of-concept and determining the functionality of a prototype based on MAS that was in compliance with a modular manufacturing line developed by us.
On Collaborative Planning for Time-constrained Multi-agent System