MPSC framework for drone traffic combines optimal control with safety supervision

In a new arXiv preprint (arXiv:2608.18353), researchers Matheus P. Loures, Guilherme V. Raffo, and Patrícia N. Pena introduce MPSC (Model Predictive Supervisory Control), a control framework designed for multi-agent systems that share resources—such as urban airspace. MPSC fuses receding-horizon cost-optimal control, the same technique used in autonomous vehicle trajectory planning, with Supervisory Control Theory (SCT), a formal methodology for enforcing safety and liveness properties. The supervision layer ensures agents never collide, never deadlock (nonblockingness), and never claim the same resource simultaneously. The key scalability breakthrough is hierarchical automaton templates: supervisors are composed locally from reusable patterns, eliminating the need for a monolithic synthesis that would blow up exponentially with the number of agents.

The authors apply MPSC to a UAS Traffic Management (UTM) model tailored for urban drone delivery. The model supports pickup-and-delivery missions under time-varying demand, a realistic scenario where request rates fluctuate throughout the day. Because the supervisor is distributed, each drone or local node runs its own safety logic while still coordinating globally—avoiding a central bottleneck. This makes the approach attractive for fleet operators and smart-city infrastructure, where centralizing all decision-making would be impractical. The 6-page paper, submitted to an IFAC conference, targets both control engineering and multi-agent systems communities. While still early-stage, MPSC offers a concrete path toward safe, scalable drone traffic management, particularly as urban air mobility grows and delivery drones become more common. Future work will likely extend the framework to larger fleets and real-world flight tests.