An agent-based approach to evaluate collaborative strategies in milk-run OEM operations

Abstract

The management of transport operations is a promising field for agent-based applications. Acting in a collaborative way, the agents can be seen as a multi-agent system that enables vehicles to handle new information perceived during the operation along cargo collecting routes. Traditional research in this field usually focus on the servicing of new transportation demands, in this paper the agents could detect the presence of unexpected traffic congestions and deciding about the transference of tasks to others vehicles. This article evaluates collaborative strategies for applying the agent-based approach to a dynamic vehicle routing formulation, where vehicles are assigned to picking-up components and parts from Original Equipment Manufacturer suppliers. A sequential probabilistic ratio test is used to predict eventual unperformed collecting tasks. Two evaluating criteria are simultaneously used to select a specific collaborative strategy: the service level, represented by the average percentage of performed tasks, plus the average daily operating cost of the vehicle fleet. The evaluation of collaboration strategies between vehicles allowed understand the functioning of distributed decision making, contributing to the adoption of autonomous control methods of transportation operations.