Abstract While learning‐based controllers demonstrate impressive performance, a significant challenge remains the lack of safety guarantees. Existing safety approaches often involve separating the optimization of the objective and constraint certification, which introduces a Model Predictive Safety Certification (MPSC) scheme. Unfortunately, this isolation frequently compromises performance, even though it ensures constraint satisfaction. This paper introduces a novel Multi‐Step MPSC scheme that combines Predictive Functional Control to track the reference output generated by the learning‐based controller, along with the utilization of a robust optimizer. The goal of this innovation is to introduce a method that confirms the system's safety while enhancing the performance of the MPSC. A numerical case study is provided and compared with the studies conducted by other researchers. The simulation results validate the efficiency and robustness of the proposed approach.
Bastani et al. (2026) studied this question.