April 11, 2025, Webb Hall 1100

Data-driven methods have become powerful tools for tackling increasingly complex problems in Systems and Control. However, deploying these methods in real-world settings — especially safety-critical ones — requires rigorous safety and performance guarantees. This need has motivated much recent work at the interface of Statistical Learning and Control, aiming to integrate formal guarantees with data-driven design methods. However, many existing approaches achieve this only by sacrificing valuable data for testing/calibration or by restricting the design space, thus leading to suboptimal performances. Against this backdrop, in this talk I will introduce Pick-to-Learn (P2L) for Systems and Control, a novel framework designed to equip any data-driven control method with state-of-the-art safety and performance guarantees. Crucially, P2L enables the use of all available data to jointly synthesize and certify the design, eliminating the need to set aside data for calibration or validation purposes. I will then demonstrate how, as a result, P2L delivers designs and certificates that outperforms existing methods across a range of core problems including optimal control, reachability analysis, safe synthesis, and robust control.