Agent-based Model for Transport Order Assignment in AGV Systems

Abstract

Transportation to solve intralogistics is becoming as complex as managing transportation in road logistics. Luckily, it takes place in structured environments where automation is easier. This work is about solving the task assignment problem for a group of automated guided vehicles (AGVs) serving the internal transport of a warehouse or factory. Instead of having a centralized task planner, we use an agent-based approach where agents represent all the stakeholders in the transport system. Namely, the clients are the transport orders and the taxis, the AGVs. We have modeled client and taxi behaviors by using extended finite-state stack machines (EFS2Ms) because they enable both modeling belief-desire-intention (BDI) agents and lower-level controllers. As a result, agent software is produced in a systematic way and, what is more, analyses of different working conditions can be done to fine-tune parameters of the models to achieve an efficient transportation. Results on one realistic study-case show that average service times can be shortened with respect to fixed planning and that this system can operate in legacy systems.