项目名称: 基于自适应神经网络的小型无人机高精度控制方法研究
项目编号: No.61273033
项目类型: 面上项目
立项/批准年度: 2013
项目学科: 自动化技术、计算机技术
项目作者: 雷旭升
作者单位: 北京航空航天大学
项目金额: 80万元
中文摘要: 复杂环境下的高精度控制是小型无人机自主性的重要标志与技术瓶颈,随着小型无人机应用需求的增加,对该问题的研究也日益迫切。针对小型无人机由于小尺寸、低速度和易形变导致的模型精度低问题,拟通过对动力学模型参数解析,基于无迹卡尔曼滤波在线辨识方法对风洞试验难以精确测量的动气动导数进行在线估计,构建高精度的小型无人机动力学模型;针对小型无人机在飞行过程中存在的模型误差、参数不确定性、外界干扰、测量误差导致的控制精度问题,拟通过构建无需样本训练的自适应神经网络系统,基于状态误差信息,在线更新权值实现对多源扰动的全包络曲线的快速估计和抑制;并针对小型无人机核心部件微导航控制系统体积小、重量轻、精度高的需求,研究基于球面拓扑3D 构型的微惯性测量单元减少非对称误差,采用干扰观测器估计并抵消测量单元存在的多类干扰,基于嵌入式信息处理单元集成设计技术,设计性能超过国际通用产品MP2028的微导航控制系统。
中文关键词: 小型无人机;系统辨识;扰动估计;自适应神经网络;自主导航系统
英文摘要: The high performance control for small unmanned aerial vehicle (SUAV) under complex environment is the important symbol and technical bottleneck. With the high requirement for small unmanned aerial vehicle, the research for this problem has become urgently. With the low size, slow speed and deformation, there exists low accuracy for the dynamic model of SUAV. This project proposes an unscented kalman filter (UKF) method to estimate the moving aerodynamic derivative parameters of SUAV in real time that wind tunnel test can not provide exactly. Focusing on the dynamic model error, parameter uncertainty, environment disturbance, the adaptive neural network without training is proposed to realize high estimation and elimination in envelope curve by updating weights in real time. Furthermore, to satisfy the requirement for low size, small weight, and high performance, a micro inertial measurement unit (MIMU) based on the 3D configuration of the spherical topology is developed to reduce the asymmetric error, and a disturbance observer is constructed to estimate and eliminate disturbance. Thus, with the embedded information unit integrated design technology, a micro guidance navigation control system can be constructed that has better performance than the international generic product MP2028.
英文关键词: small unmanned aerial vehicle;system identification;distrubance estimation;adaptive neural network;autonomous navigation system