Data-Driven Safety-Critical Control with High-Order Iterative Control Barrier Function

The stability and safety of uncertain nonlinear systems have always been an essential and hard task in the automation field. For the forward invariance of the safety set of the system under unmodeled dynamics, a new concept of data-driven high-order iterative control barrier function (DHICBF) is presented. To overcome the structural complexity and dependence on dynamic models of traditional robot controllers, an iterative learning mechanism is introduced and a modelfree iterative predictive controller (MFIPC) is designed. The controller can be progressively optimized by predicting future information so that it can cope with dynamic changes and uncertainties in the environment. On this basis, quadratic programming (QP) that unifies the DHI-CBF with the MFIPC is established, which is able to prioritize the safety of the system in case of conflict between the desired output and the security boundary. Finally, the application for Franka-Panda robot demonstrates the superiority of the presented algorithm.