【作者】 蒋鸿翔；

【导师】 高正； 徐锦法；

【作者基本信息】 南京航空航天大学 ， 飞行器设计， 2008， 博士

【摘要】 本文所研究的新型无人直升机(NUH)具有不同于常规直升机的机身结构与气动布局,将在许多应用场合中发挥重要的作用。利用计算机视觉实现自动进场着陆,将显著节省NUH着陆所消耗的人力和物力,具有精度高、功耗低、安全可靠等优点。研究NUH视觉导引着陆这一无人机导航与控制新技术将具有重大的应用价值。本文提出了基于视觉的进场着陆方案,将NUH着陆过程分为导引进场与落地(着陆)两阶段,研究内容涉及NUH数学建模、着陆控制设计以及每一阶段所需视觉算法设计和验证等关键技术。为满足NUH视觉导引着陆控制需要,利用机理建模方法建立NUH非线性模型,分析了NUH飞行动力学特性,完成了悬停点配平计算,给出了配平点的小扰动线性模型,完成了非线性模型和线性模型的操纵响应对比分析,明确了应用线性模型可以完成NUH视觉导引着陆控制设计研究。NUH具有开环不稳定性和独特的操纵响应耦合特性,利用特征结构配置方法进行了NUH悬停增稳设计,分析了不同的特征结构配置对NUH闭环系统稳定性与操纵性的影响。运用内外回路控制设计思想,实现了NUH悬停着陆控制系统。内回路中利用特征结构配置实现角速度与垂向速度的稳定与控制。姿态控制回路由内回路扩展而成,内回路输出信号的积分作为姿态角反馈信号,与姿态角指令信号比较后得到误差信号,经比例控制器作用后作为内回路的输入。线速度与线位移控制回路同样由上述方法扩展而得。仿真结果表明,内回路可实现快速指令跟踪和响应解耦,外回路可实现悬停位置保持和着陆轨迹跟踪。提出NUH进场阶段的视觉导引原理方案,进行相应的视觉导引算法设计。运用图像处理算法得到了导引标志图像的中心点位置和面积,作为视觉测量值,运用扩展卡尔曼滤波算法实现视觉与惯性数据融合,由视觉测量值修正惯性数据递推估计结果,得到更为精确的NUH位置估计值。NUH位置控制回路利用这一估计值作为反馈量,实现NUH进场导引的位置反馈控制。视觉在回路导引仿真验证了NUH视觉导引方案的可行性。提出NUH落地阶段的视觉着陆原理方案,进行相应的视觉着陆算法设计。运用图像处理算法得到着陆标志特征角点在不同图像中的像点位置与像点速度,利用点对应关系估计摄像机坐标系之间的运动,从而得到NUH的角速度、姿态、线速度以及NUH相对着陆平台的位置估计。运动参数估计值作为内回路的状态反馈量和外回路的姿态、线速度和位置反馈量,实现NUH悬停着陆控制。视觉在回路着陆仿真验证了NUH视觉着陆方案的可行性。为验证视觉导引着陆算法和控制算法,设计了小车运动试验验证方案。阐述了小车试验平台的研制,包括小车本体、定位定向设备、控制系统和视觉系统的设计与实现。进行小车运动试验,证实了视觉系统能有效地执行NUH导引着陆过程中的图像数据提供与处理。更多还原

【Abstract】 Novel unmanned helicopter (NUH) owns special airframe structure and aerodynamic configuration that are different from conventional helicopter’s and will play an important role in many application situations. As soon as NUH can automatically approach and land to the landing pad using computer vision, the manpower and material resource consumptions can be significantly reduced. This method’s merits include high precision, low power dissipation, safety and reliability. Vision-based guidance and landing technology is a new flight control and navigation technology of unmanned helicopter, the research on it has great application value. According to this target, a vision-based guidance and landing control scheme was proposed, in which the NUH automatic approaching and landing was divided into the guidance stage and the landing stage. Main research contents included some key technologies such as NUH mathematics modeling, landing control, design and verification of principle of vision algorithm at each stage.To meet the necessity of NUH vision-based guidance and landing control, a NUH nonlinear model was built using aerodynamic principle modeling method. NUH flight dynamics characteristics were analyzed. The NUH trim calculation in hover condition was completed. A NUH linear model was deduced using linearizing method under the assumption of small perturbation. A comparison analysis about control responses of nonlinear and linear model was conducted. It was demonstrated by simulation that the linear model can be applied to NUH vision-based guidance and landing control design.NUH has open-loop instability and special control response coupling. NUH hover stability augment system was designed using EA method. The effects on NUH close-loop stability and maneuverability with different eigen-structure configurations were discussed. NUH hover/landing control system was designed using EA method combined with inner-outer-loop control architecture. The inner control loops in which the stabilizing and control of angular rates and vertical velocity were implemented were built using EA method. The attitude angular control loops were constructed by extending the inner loops: the outputs of the inner loops were integrated first; then as feedback signals, the integrated outputs were compared with the setting attitude angular signals; eventually the attitude angular error signals were amplified with proportional controllers and sent to the inner control loops as input commands. The control loops of linear velocity and position were implemented similarly using above method. It was demonstrated by NUH hover/landing simulation that the designed inner control loops can realize well decoupled and quickly angular rate commands tracking, and that the outer control loops have good hover position hold and landing trajectory following abilities.A vision-based guidance principle scheme for the NUH approaching stage was proposed. A corresponding VGA was designed. The central point position and the size of the guidance target were decided and calculated using image processing. These were treated as vision measured data. The inertial data and the vision data were fused using EKF formulation in which the estimated result from inertial data was revised by vision measurement. Then more precise estimated NUH positions were calculated. Using the estimated position as feedback, the position control loops carried out NUH to approach the landing pad. Feasibility of vision-based guidance scheme was demonstrated with vision-in-loop guidance simulation.A vision-based landing principle scheme for the NUH landing stage was proposed. A corresponding VLA was designed. The image position and velocity of corner points of the landing target in different images were calculated using image processing. The motion parameters of camera were estimated using the correspondences of corner points in different images. The estimated angular rates, attitude angles, velocities and the relative position to landing target of NUH were get. Treating the estimated motion parameters of NUH as state feedbacks for inner control loop and as attitude, velocity and displacement feedbacks for outer control loop, hover/landing control was implemented. Feasibility of vision-based landing scheme was demonstrated by vision-in-loop landing simulation.To verify the vision-based guidance, landing and control algorithm, a scheme of special moving platform experiment was proposed. The design and implement of components of the platform, such as the platform frame, the orienting and locating devices, the moving control system and the embedded vision system, were described. By the ground experiment, it was demonstrated that the designed embedded vision system can effectively execute the image acquisition and processing during NUH vision-based guidance and landing.更多还原