Research & Papers

New supervisory control method introduces multimodal nonblocking for safer systems

A novel nonblocking variant tackles fault behavior and reconfiguration in critical infrastructure.

Deep Dive

A team led by Marijn Minkenberg (Eindhoven University of Technology) has proposed a new approach to supervisory control synthesis that addresses a key limitation: the standard nonblocking property is too weak for systems that include fault behavior, reconfiguration, or multiple control goals. Their paper, submitted to the 2026 IEEE Conference on Decision and Control, introduces 'modal' and 'multimodal nonblocking' variants. These impose targeted restrictions on the states visited along paths to marked states, ensuring liveness is more robust in complex scenarios. The authors present concrete synthesis algorithms to construct supervisors that guarantee these new nonblocking properties.

The practical value is demonstrated through three intuitive examples inspired by real-world safety-critical water infrastructures. The team shows that their novel nonblocking variants are distinct from established ones, and they formulate conditions under which one variant implies another. This work is particularly relevant for engineers designing control systems for water treatment plants, power grids, or autonomous factories where failures and reconfiguration are common. By offering stronger liveness guarantees, the method could reduce the risk of deadlocks or unresponsive states in automated safety-critical systems.

Key Points
  • Introduces modal and multimodal nonblocking as stronger liveness properties for supervisory control systems.
  • Provides synthesis algorithms that construct supervisors guaranteeing these new nonblocking variants.
  • Validated on three examples from safety-critical water infrastructures, showing distinctness from existing nonblocking notions.

Why It Matters

This strengthens liveness guarantees in automated systems with faults and reconfiguration, improving safety in critical infrastructure.

📬 Get the top 10 AI stories daily