Title:Robust Finite-Time Adaptation Law for Safety-Critical Control

Abstract:The process of parameter adaptation and identification relies on the excitation of the measurable signals. To ensure at least interval excitation, designed control or noise input is often introduced. The main concern in this paper is the safety of adaptation process, i.e. the empirically chosen noise input may lead system state into unsafe region. First, a safe adaptation framework is presented to achieve both finite-time parameter identification and guaranteed safety. The element-wise parameter estimation is obtained using finite-time dynamic regressor extension and mixing (FT-DREM), by which a worst-case estimation error can also be computed. In this situation, a point-wise safe strategy based on adaptive control barrier function (aCBF) is used to filter the designed control or noise input. Some criteria are proposed to reduce the conservativeness of aCBF method. The robustness of the proposed algorithms is analyzed and a robust algorithm for safe adaptation is also presented. In the end, a simulation experiment of adaptive cruise control with estimation for slope resistance is studied to illustrate the effectiveness of the proposed methods.