Automated Guided Vehicles (AGVs) are among the most popular automated transportation systems in flexible manufacturing systems (FMSs). Bi-directional AGVs are particularly efficient; nevertheless, their control is complex as they require effective routing procedures to avoid many undesired situations such as head-on conflicts and deadlocks. This strongly limits their use in production and storage facilities, especially due to the lack of easy-to-implement control procedures. This paper proposes a new zone-control architecture of bi-directional AGVs based on two stages. Any route-computation method can be used in the first stage to minimise some performance criterion. The paper focuses on the second stage where the problem is formalised based on supervisory control theory (SCT) in order to secure AGVs paths. This hybrid architecture provides state-space reduction compared to similar SCT-based methods as demonstrated in the simulation study. It also ensures robust deadlocks-prevention in presence of disturbances in case AGVs are scheduled to avoid them.