SINDy-CBF: Data-Driven Identification and Safe Control of a Planar Quadrotor

Quadrotors in dynamic environments require precise modeling and safety guarantees. Sparse Identification of Nonlinear Dynamical systems (SINDy) offers an effective method for learning precise quadrotor dynamics, enabling improved control. While model-based control has achieved breakthroughs, ensuring safety with dynamic obstacles remains challenging. We propose a control framework that considers the nonlinear, underactuated dynamics of a planar quadrotor identified via SINDy. A differential flatness-based controller is developed with a Control Barrier Function (CBF) filter to enforce safety. Specifically, we use an Augmented Control Barrier Function (ACBF) to define a safety region in position space while accounting for the quadrotor’s orientation, thereby constraining the control moments. Our numerical results validate the proposed controller through scenarios including set point tracking, trajectory tracking with static obstacles, and trajectory tracking with dynamic obstacles.