障碍物
计算机科学
控制工程
控制(管理)
控制理论(社会学)
控制系统
工程类
雷达跟踪器
自动控制
弹道
鲁棒控制
过程控制
车辆动力学
自动化
稳健性(进化)
矩阵代数
加速度
避障
信号处理
概率密度函数
最优控制
作者
Ang Li,Hongtao Yin,Ping Fu,Lan Duo
标识
DOI:10.1109/taes.2026.3729536
摘要
Considering complex obstacle environments, we propose a safety-critical formation control method for quadrotor unmanned aerial vehicles (QUAVs) that ensures high-probability safety. The main challenge is that, in complex obstacle environments, the safety of QUAVs cannot be directly characterized by deterministic safety functions. Instead, it is typically inferred from safety samples and estimated global safety functions, which poses challenges in evaluating safety probabilities and guaranteeing high-probability safety. To address this, we introduce a safety threshold mapping function (STMF) to obtain safety samples. Meanwhile, to address model uncertainties and external disturbances of QUAVs, we employ a Gaussian process regression observer (GPRO) for online estimation. Then, we propose a dual Gaussian robust control barrier function (DGR-CBF), which uses the safety samples generated by STMF as the training set and employs Gaussian process regression (GPR) to estimate the safety function. By fusing safety belief and safety uncertainty, the proposed method ensures high-probability system safety while providing an explicit safety probability. Similarly, a robust control barrier function (RCBF) is designed by incorporating the compensation results from GPRO to ensure collision avoidance between QUAVs. Finally, we design a safety-critical quadratic programming framework that employs the formation nominal controller as the objective function and enforces the aforementioned control barrier functions (CBFs) as hard constraints. The comparative simulation experiments demonstrate the advantages of the proposed method in formation accuracy, high-probability safety, and robustness.
科研通智能强力驱动
Strongly Powered by AbleSci AI