A Machine Learning Approach to Passivity-Preserving Safety-Critical Control

• Passivity, a property of systems that describes their ability to store and dissipate energy, is important to ensure stability and robustness. Control barrier functions are used to ensure that the system remains within a safe set of states. The paper proposes a new machine-learning-based method for the synthesis of control barrier functions that preserve the passivity of safety-critical controlled systems, also by accounting for input saturation. • The effectiveness of the proposed control strategy is validated through numerical simulations on a trolley-pole system and a 2R robot. The simulations show that the proposed method is capable of preserving passivity and ensuring the safety of the system, while respecting the input constraints. Passivity-Based Control enables to constructively synthetize controllers with guarantees on the stability of the overall system. Although safety-critical control based on Control Barrier Functions is able to synthetize controllers that effectively cope with state constraints, passivity and input bounds are not addressed in general. This paper proposes a novel model-based Machine Learning methodology aimed at synthesizing Control Barrier Functions such that passivity of the closed loop system is preserved under safety-critical control and input saturation. Numerical simulations on a cart-pole system and on a 2R robot validate the effectiveness of the proposed control strategy in terms of performances and passivity preservation.