【作者】 张驰；

【导师】 张涛；

【作者基本信息】 东南大学 ， 精密仪器及机械， 2023， 硕士

【摘要】 随着全球经济的发展和各国对海洋安全的重视,海洋环境探测和资源开采成为了当前各国研究的热门课题。水下航行器在这些海洋活动中发挥着不可替代的作用,准确的导航定位信息是水下航行器正常工作的前提条件。受到水下环境的影响,传统陆上导航方法无法使用。惯性导航成为水下航行器的主要导航方式,但是由于其本身器件的原因,导航误差会随时间逐渐累加,无法满足高自主、长航时的水下导航需求。为提升水下导航定位精度,本文研究基于多波束测深的地形匹配导航算法,论文主要工作如下:(1)讨论了水下地形匹配导航系统的构成单元,对单波束测深与多波束测深的原理进行了分析比较,介绍了水下地形图的数学模型,并给出了地形图网格内部的水深插值方法,比较了TERCOM、SITAN和ICCP算法的原理,对捷联惯性导航系统的原理和误差模型进行了介绍。(2)针对多波束原始数据处理问题进行研究。提出了一种基于DBSCAN聚类的多波束异常值识别方法,减小异常值对后续地形匹配的影响。为减少地形匹配算法的计算量,提出了一种基于自适应阈值和距离约束的多波束数据抽稀方法。实验结果表明,本文提出的基于DBSCAN聚类的异常值识别方法能够有效检测出异常数据,相比于其他算法,本文提出的抽稀算法能够更好保持地形原始特征。(3)针对传统水下地形匹配算法的不足,提出了一种基于改进遗传算法的水下地形匹配方法。该方法在多波束测深数据中选择多条水深序列,以遗传算法为基础进行地形匹配,对于传统遗传算法可能陷入局部最优的不足,提出使用正余弦算法对其进行优化,提高算法收敛速度,改善局部收敛的问题。实验结果表明,所提方法对初始位置误差的大小不敏感,平均匹配精度在3m以内。(4)针对连续导航的需求,搭建了基于粒子滤波的惯性/地形松组合导航模型。在相似地形处,粒子滤波的似然函数容易出现多峰问题,对此提出了一种多峰检测方式,在检测到多峰后进行权值回溯处理,以减小多峰对位置估计的影响。船载实验表明,本文所提算法水平位置的平均误差约为2m,最大误差小于7m,导航定位精度得到提高。更多还原

【Abstract】 As the global economy develops and countries place greater emphasis on maritime security,marine environment surveying and resource exploration have become hot topics within various nations’ research agendas.Underwater vehicles play an irreplaceable role in these marine activities,and accurate navigation and positioning information are prerequisite for the normal operation of underwater vehicles.Due to the effects of underwater environment,traditional land-based navigation methods cannot be used.Inertial navigation has become the main navigation method for underwater vehicles.However,due to the inherent characteristics of its devices,navigation errors will gradually accumulate over time,which cannot meet the requirements of high autonomy and long endurance requirements of underwater navigation.To improve the accuracy of navigation and positioning,a terrain matching navigation algorithm based on multibeam bathymetry is studied.The main contributions of the paper are as follows:(1)The constituent units of underwater terrain matching navigation system are discussed,and an analysis and comparison of the principles between single-beam and multibeam is made.Moreover,the mathematical model of underwater terrain maps is introduced,and a method is presented to interpolate the water depth inside the grid of terrain maps.Furthermore,the principles of TERCOM,SITAN,and ICCP algorithms are compared.Additionally,the principle and error model of strapdown inertial navigation systems are introduced.(2)The research on the processing of multibeam raw data is carried out.A multibeam anomaly detection method based on DBSCAN is proposed to reduce the impact of anomalies on subsequent terrain matching.In order to reduce the computational complexity of the terrain matching algorithm,a multibeam data thinning method based on adaptive threshold and distance constraint is proposed.The experimental results show that the anomaly recognition method based on DBSCAN can effectively detect abnormal data,and the thinning algorithm can better preserve the original features of the terrain when compared to other algorithms.(3)To address the shortcomings of traditional terrain matching algorithms,a new underwater terrain matching method based on an improved genetic algorithm is proposed.This method selects multiple water depth sequences from multibeam data and uses a genetic algorithm for terrain matching.To address the problem that the traditional genetic algorithm may fall into local optimum,the proposal suggests optimizing it using the sine and cosine algorithm,which improves the convergence speed of the algorithm and solves the problem of local convergence.Experimental results indicate that the proposed method is insensitive to initial position errors and the average matching accuracy is within 3m.(4)To meet the requirements of continuous navigation,an inertial/terrain loosely coupled navigation model based on particle filtering is established.In areas with similar terrain,the likelihood function of particle filtering is prone to have multi-peak problems.To address this issue,a multi-peak detection method is proposed,and weight backtracking processing is performed after multiple peaks are detected to reduce the impact of multi-peaks on position estimation.Shipboard experiments show that the proposed algorithm achieves an average error of approximately 2m in horizontal position,with a maximum error of less than 7m,and the precision of navigation and positioning is improved.更多还原