A Local-observation-based Safe Autonomous Navigation Approach in an Unknown Environment

Control Lyapunov functions (CLFs) and control barrier functions (CBFs), serving as constraints in quadratic programming (QP), have been demonstrated to provide safe navigation using a specified map. However, the practical implementation of CBF-CLF-QP for autonomous navigation in unknown environments faces the challenge of formulating the convex CBFs using sensory data. This paper presents a pipeline to construct the local control barrier functions based on observed point cloud data and forming a convex optimization problem. We use iterative plane fitting and cuboid bounding boxes to estimate the CBFs, which can separate safe and unsafe regions and ensure that the optimization problem is convex. The proposed method achieves mapless safe autonomous navigation in 3-dimensional space, which has the potential to enhance the level of autonomy for unmanned aerial vehicles (UAVs). The effectiveness of the proposed method is validated via numerical simulations.