【作者】 张大兴；

【导师】 贾建援；

【作者基本信息】 西安电子科技大学 ， 机械制造及其自动化， 2008， 博士

【摘要】 光电捕获、跟踪与瞄准(Acquisition,Tracking and Pointing,ATP)系统以高于微波频率的光波为信号载体,使跟瞄系统战术性能产生了质的飞跃,其在武器控制、天文观测、靶场测量、航空航天和激光通信等领域取得了日益广泛的应用。虽然在不同领域对ATP系统要求侧重点各不相同,但总的趋势都是向高精度、高速度、强适应性发展。随着对光电ATP系统要求的不断提高,特别是光电ATP系统自身机动性的要求以及为避开大气对目标观测或者光束传输的影响等原因,基座固定于大地的光电ATP系统已经不能满足使用要求,发展基座固定于运动平台上的光电ATP系统成为当前研究的重要内容。但是相对于地基ATP系统,运动平台ATP系统面临更多问题。要达到高精度、高速度、强适应性,除了要有高性能的光电传感元件之外,快速、灵活而精度高的伺服系统也是关键。跟瞄装置执行机构的结构形式、传动装置存在的摩擦、齿隙等非线性的影响等是制约伺服系统精度进一步提高的重要因素。另外,随着现代战争、航天航空、检测技术、智能控制等领域技术的不断进步,无人驾驶、智能避障、路径智能规划的自动化运动平台,也成为运动平台ATP系统研究的一个重要领域。基于以上现状,本文对三轴跟踪架ATP运动平台系统中的跟瞄特性、三轴跟踪架结构的动力学建模、三轴跟瞄策略优化、齿隙摩擦补偿、路径规划等方面进行了研究:基于自主设计制造的运动平台ATP系统,介绍了三轴跟踪架结构存在冗余自由度、跟瞄策略多样化的特点,分析了偏距对系统跟瞄性能的影响,指出了偏距引入的不可见区域以及误差的处理方法。基于刚体定轴转动的四元数描述方法,采用Lagrange-Maxwell方程,建立了三轴跟踪架结构ATP系统的机电动力学模型。基于建立的动力学模型,分析了三轴ATP系统的运动形式,仿真和实验结果基本一致,验证了该动力学模型的正确性。将非线性权值递减策略引入改进的Meta粒子群(M2PSO)算法,提出了一种改进的M2PSO算法,采用经典函数Sphere和Rosnbrock测试了M2PSO算法和改进M2PSO算法的寻优性能,测试结果表明,改进M2PSO算法较M2PSO算法有更快的收敛速度。通过分析三轴ATP系统的跟瞄运动过程,在给定当前三轴位置角和探测器误差角的条件下,将求解三轴角增量的方法转化为求最优值问题,并采用改进的M2PSO算法求得了不同初始条件下满足目标要求的三轴角增量组合,通过与方位俯仰两轴结构的仿真结果比较,验证了三轴跟踪架系统全空间无跟踪盲区的优点。将RBF神经网络引入到滑模变结构控制中,阐述了RBF神经网络滑模变结构复合控制器的设计过程,并对复合控制器进行了稳定性分析,提出了满足控制器稳定性要求的条件。复合控制方法采用RBF网络的在线学习功能,调节滑模变结构控制切换项的增益,能降低滑模变结构控制的抖振现象。将集合摩擦模型和滞环齿隙模型引入三轴传动系统机电动力学模型,建立了同时考虑摩擦、齿隙以及传动轴刚度变化特性等因素的系统动力学模型。基于以上模型,采用RBF神经网络滑模变结构复合控制方法对系统存在的齿隙和摩擦现象进行了补偿控制。仿真比较了PD控制、固定增益滑模变结构控制和RBF神经网络滑模变结构控制三种方法的控制效果,仿真结果表明RBF神经网络滑模变结构控制的补偿效果优于PD控制和固定增益滑模变结构控制,是一种有效的齿隙和摩擦补偿方法。建立了运动平台小车的运动学和动力学数学模型,提出了一种基于传感器检测实时更新的动态环境地图的栅格法描述方法。针对传统人工势场法存在局部最小值的问题,采用一种基于PSO算法的改进人工势场法,给出了改进方法的实现流程,仿真结果表明,改进的人工势场法能够解决传统人工势场法存在的局部最小值问题,是一种有效的路径规划方法。更多还原

【Abstract】 The signal of photoelectric ATP (Acquisition,Tracking and Pointing) system is carried by light-wave, whose frequency is higher than microwave. Therefore, the tactical performance of ATP system was constitutionally improved. The applications in arms control, celestial observation, shooting range determine, aeronautics and astronautics, laser communication and etc. were constantly extended. The performance demand of ATP system is dissimilar in different fields, but the trend of development is higher precision, faster speed and more extensive applicability.Because of the continuously demand for higher performance of photoelectric ATP system, especially for the maneuverability, keeping away from atmosphere and light transfers and etc. the photoelectric ATP system based on the earth can not meet the demand. The development of photoelectric ATP system setup on moving platform becomes the significant question for discussion recently.Comparing with the ATP system based on the earth, the ATP system setup on moving platform has more questions to deal with. Arm to achieve higher precision, faster speed and more extensive applicability, besides the high performance photoelectric transducer, the servo system providing with fast speed, flexibility and high precision is also the key question. The configuration of executive machine, the backlash and friction consisting in gearing are all the primary factors limiting to improve the precision of servo system. Otherwise, along with the development of modern war, aeronautics and astronautics, detection technology and intelligent control, the moving bed of unpiloted, avoiding collision intelligently and self-acting path planning becomes one of the dominant fields of ATP system.Base on the above perception, the tracking and pointing characteristic of three-axis ATP system, the dynamic modeling of three-axis structure, the three-axis optimization of tracking and pointing, backlash and friction compensation and path planning and etc. were investigated as follows.Base on the ATP system setup on moving platform designed all by us, the multifarious characteristic of tracking and pointing strategy was introduced. The effects of deflection distance on tracking capability were analyzed, and the methods deal with the sightless area and error caused by deflection distance were introduced.The dynamic model of three-axis ATP system was established base on Lagrange-Maxwell functions and the quaternary method for describing the rigid body rotation. The movement of three-axis ATP system was analyzed based on the dynamic model. The validity of this dynamic model was validated by simulation and experiment results.The idea of nonlinear decreasing inertia weight strategy was introduced in modified Meta particle swarm optimization (M2PSO), and an improved M2PSO was presented. The performance of M2PSO and improved M2PSO was tested by Sphere and Rosnbrock functions. The test results suggested that the convergent speed of improved M2PSO is higher than that of M2PSO.Based on the movement analysis of three-axis ATP system, the solution of angle increment was transformed to a question of optimization. If the current angles and the detect errors are known, the angle increment of three-axis can be solved by improved M2PSO.A method of RBF neural network sliding mode variable structure compound control was presented. The design process of compound controller was introduced in detail. The stability of compound controller was analyzed, and the stable condition of compound controller was put forward. The compound control can reduce the buffeting of sliding mode variable structure control by adopting the learning function online of RBF neural network, which can adjust the switch gain of sliding mode variable structure control.The dynamic model of three-axis drive system including backlash and friction was established based on the backlash hysteresis model and friction aggregation model. The RBF neural network sliding mode variable structure compound control was adopted as the backlash and friction compensate control method. The effectiveness of PD control, sliding mode variable structure control and RBF neural network sliding mode variable structure control were simulated and compared. Results suggest that RBF neural network sliding mode variable structure control can achieve higher precision than PD control and sliding mode variable structure control.The kinematic and dynamic model of the moving platform was established, and a grid method for uncertain environments based on sensors was presented. Because the classical potential field method has the problem of local minimum, a modified potential field method base on PSO was put forward. The program flow of improved method was presented in detail. Simulation results suggest that the modified potential field method can avoid the local minimum of classical potential field method.更多还原