基于数据驱动模型预测控制的无人机轨迹跟踪方法  

Data-driven model predictive control based method for unmanned aerial vehicle trajectory tracking

作　　者：於怿丰 任思维 张鑫帅 谢芳芳 季廷炜[1] 杜昌平[1] YU Yifeng;REN Siwei;ZHANG Xinshuai;XIE Fangfang;JI Tingwei;DU Changping(School of Aeronautics and Astronautics,Zhejiang University,Hangzhou 310058,China)

机构地区：[1]浙江大学航空航天学院,杭州310058

出　　处：《兵器装备工程学报》2024年第11期272-282,共11页Journal of Ordnance Equipment Engineering

摘　　要：提出一种针对固定翼无人机自适应控制行为的新策略,用于在复杂多变工况下更好地完成轨迹跟踪任务。该方法通过小规模实时数据驱动在线稀疏辨识,构建和更新非线性动力学系统模型并应用于模型预测控制器,以快速适应不同工况。飞行试验结果表明,动态更新模型参数能够带来更优的控制性能,在复杂有风与无风的轨迹跟踪任务中,数据驱动与在线辨识的模型预测控制器都能够使得无人机的飞行轨迹与目标路径保持良好的趋同;微型机载计算机的实验证明,本方法产生的计算开销合理,可适用于大部分边缘计算平台。A new strategy is proposed for adaptive behavior control of fixed-wing drones,aiming at enhancing performance of trajectory tracking tasks under complex and varied conditions.This method utilizes real-time data and employs the online sparse recognition to construct and update a nonlinear dynamic system model.By applying this updated model to predictive controllers,it can swiftly adapt to varying circumstances.Flight test results show improved control performance through dynamic updates of the model parameters.In both windy and wind-free conditions,the data-driven predictive controller with online recognition can keep the drone’s flight path aligning closely with the target route.An empirical demonstration on a miniaturized onboard computer confirms that computational overhead of this method stays within a reasonable range,making it suitable for most edge computing platforms.

关 键 词：数据驱动 在线辨识 模型预测控制 固定翼无人机 轨迹跟踪 稀疏识别 

分 类 号：V249[航空宇航科学与技术—飞行器设计]