Set-Based Control Barrier Functions and Safety Filters

Abstract: High performance and formal safety guarantees are common requirements for industrial control applications. Control barrier function (CBF) methods provide a systematic approach to the modularization of safety and performance. However, the design of such CBFs can be challenging, which limits their applicability to large-scale or data-driven systems. This paper introduces the concept of a set-based CBF for linear systems with convex constraints. By leveraging control invariant sets from reachability analysis and predictive control, the set-based CBF is defined implicitly through the minimal scaling of such a set to contain the current system state. This approach enables the development of implicit, data-driven, and high-dimensional CBF representations. The paper demonstrates the design of a safety filter using set-based CBFs, which is suitable for real-time implementations and learning-based approximations to reduce online computational demands. The effectiveness of the method is illustrated through comprehensive simulations on a high-dimensional mass-spring-damper system and a motion control task, and it is validated experimentally using an electric drive application with short sampling times, highlighting its practical benefits for safety-critical control.