Exponential control barrier function and model predictive control for jerk-level reactive motion planning of quadrotors

Quadrotors require efficient reactive motion planning algorithms to ensure safe autonomous operations in dynamic environments. One common strategy for reactive motion planning is model predictive control (MPC); however, conventional MPC-based methods often fall short of guaranteeing collision avoidance when encountering dynamic obstacles. To address this limitation, we develop an enhanced framework that combines MPC with control barrier functions (CBFs) for improved safety and a Kalman Filter (KF) for predicting obstacle behavior, increasing responsiveness to dynamic obstacles. We utilize a high-order closed-loop model of the quadrotor along with exponential CBFs, enabling trajectory control at the jerk level, unlike existing MPC-CBF methods that rely on acceleration-level planning. Extensive hardware experiments across multiple scenarios demonstrate that this approach significantly enhances safety by increasing the minimum vehicle-obstacle distance and enabling successful navigation through complex situations, such as avoiding fast-swinging obstacles, where traditional MPC-only methods fail. Hardware-based sensitivity analysis further reveals the algorithm's overall robustness to variations in parameter values, provides insight into parameter tuning, and highlights the critical role of accurate obstacle predictions in dynamic environments. Our findings indicate that the MPC-CBF-KF framework is a promising, robust, and computationally feasible solution for quadrotor motion planning in both dynamic and static environments.