Delegate MASs for coordination and control of one-directional AGV systems: a proof-of-concept

Abstract

Decentralized coordination and route planning face the challenges such as scalability, dynamic changes (disturbances) in the environment, continuous planning, and coordination issues (i.e., deadlock and livelock situations). Self-organized delegate multi-agent systems (D-MASs) have proven to be effective decentralized coordination mechanisms for coordination and control (C&C) applications. However, the use of such coordination mechanisms becomes more challenging, compared to the previous studies, in which the coordinated entities are one-directional automated guided vehicles (AGVs), with restricted movement, situated in a highly dynamic production environment. To address these challenges, there were several problematic situations identified dealing with issues such as the originally proposed functionalities of D-MASs, restricted movement, priority parameter settings, and simulated failures of AGVs. Solutions (coordination rules) to these situations were proposed, also described examples were provided and, finally, the approach was verified by simulation in the 3D environment, involving five AGV agents (AGVAs). Simple indicators of such intralogistics system were proposed to outline the system performance. Simulations were performed with as well as without simulated failure states. Simulation results show that the proof-of-concept was reached, and that by the combination of the proposed coordination rules and D-MAS, one-directional AGVAs were able to generate a short-term forecast for the near future and thus anticipate and avoid coordination issues as well as to cope with simulated failures.