GPS信号短时缺失情况下的组合导航数据融合方法研究

发布时间：2018-08-02 16:59

【摘要】：在现代化军事战争中,对于高动态的飞机、导弹、舰船等的跟踪精度和可靠性要求越来越高,单个系统已经无法满足要求。为了解决这一问题,一般采用将两种或多种导航定位系统组合在一起进行优势互补。导航界采用较多的组合方式是将GPS和惯性导航系统结合在一起实现对载体的导航定位,在组合的过程中需要GPS对惯导进行一定频率的修正,从而限制其误差随时间的积累。但是当GPS信号受到干扰或者被遮挡就无法实现对惯导的修正。因此本文主要研究高动态运载体上的GPS/INS组合导航系统在GPS信号短时缺失情况下的数据融合方法。对于如何提高系统精度和可靠性,本文主要做了以下工作：1、以飞机作为高动态的运载体,模拟了其多个飞行状态,通过Matlab仿真产生航行轨迹。并以其位置、速度、姿态角等数据为基础,仿真INS的加速度计和陀螺仪模块,通过解算产生组合系统融合所需要的INS位置数据,与运载体真实的轨迹进行对比,验证了INS误差随时间积累的特点,因此需要GPS的数据对其进行频繁修正,显示了本文研究的必要性。2、研究人工神经网络的理论及其特点,分析标准BP神经网络在收敛速度和泛化能力方面存在的问题,针对其在GPS/INS信号处理模型应用中存在的不足,提出一种基于L-M贝叶斯正则化方法的BP神经网络INS误差预测模型,当GPS信号中断时可以利用该模型预测INS的误差来修正其位置数据。并与标准的BP神经网络预测模型得到的结果进行比较,验证其优势。3、由于BP神经网络模型初始参数一般取随机数或者根据经验给定,造成网络模型的不稳定甚至使网络陷入局部最小值。所以本文提出一种结合改进粒子群优化方法和BP神经网络的混合算法,用于建立INS误差预测模型。首先通过粒子群算法对神经网络的初始参数进行优化,然后再对BP神经网络进行训练,产生预测模型。通过Matlab仿真表明,该混合算法可以有效提高模型的收敛速度和泛化能力。
[Abstract]:In modern military warfare, the tracking accuracy and reliability of highly dynamic aircraft, missiles, ships and so on are more and more high, and a single system can not meet the requirements. In order to solve this problem, two or more navigation and positioning systems are usually combined to complement each other. In navigation field, GPS and inertial navigation system are combined to realize navigation and positioning of carrier. In the process of integration, GPS is required to modify inertial navigation in a certain frequency, thus limiting the accumulation of errors over time. But when the GPS signal is disturbed or blocked, the inertial navigation can not be corrected. Therefore, this paper mainly studies the data fusion method of GPS/INS integrated navigation system in the case of short time absence of GPS signal on high dynamic vehicle. As to how to improve the accuracy and reliability of the system, this paper mainly does the following work: 1, taking the aircraft as the high dynamic carrier, simulates its multiple flight states, and generates the flight trajectory by Matlab simulation. Based on the data of position, velocity, attitude angle and so on, the accelerometer and gyroscope module of INS are simulated, and the INS position data needed for the fusion of the combined system are generated by solving, and compared with the real trajectory of the vehicle. The characteristic of INS error accumulation with time is verified, so it needs the data of GPS to revise it frequently. It shows the necessity of the research in this paper. The theory and characteristics of artificial neural network are studied. This paper analyzes the problems in convergence speed and generalization ability of standard BP neural network, and proposes a BP neural network INS error prediction model based on L-M Bayesian regularization method in view of its shortcomings in the application of GPS/INS signal processing model. When the GPS signal is interrupted, the model can be used to predict the error of INS to correct its position data. Compared with the results obtained from the standard BP neural network prediction model, the advantages of BP neural network model are verified. 3. Because the initial parameters of BP neural network model are generally random numbers or given according to experience, It causes the instability of the network model and even makes the network fall into the local minimum value. So this paper presents a hybrid algorithm combining improved particle swarm optimization and BP neural network to establish INS error prediction model. The initial parameters of neural network are optimized by particle swarm optimization (PSO), and then BP neural network is trained to generate prediction model. The Matlab simulation shows that the hybrid algorithm can effectively improve the convergence rate and generalization ability of the model.
【学位授予单位】：西北大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：P228.4;TN96

【相似文献】