Computation of a maximally permissive controller in the Ramadge-Wonham framework promises a general solution to automatically design a controller for a discrete event system, when it exists. However, like for all similar model-checking approaches, the combinatorial explosion of the state space remains a practical issue. The work presented here investigates how to exploit both hierarchical modeling and a symbolic model-checking engine to tackle this problem. This engine is based on a powerful class of Decision Diagrams called Hierarchical Set Decision Diagrams combined with a framework called Instantiable Transition Systems, in order to describe hierarchical models. To implement the controller activity, we propose to store the set of safe states, computed offline, as a decision diagram in the controller software, allowing to take decisions on-line. We run a prototype tool on several benchmark examples, including a problem of automated guided vehicles and a train crossing version with explicit discrete time. Results suggest good scalability, although the procedure is computationally intensive.