中文摘要：

为提高弱通信条件下多UUV系统互定位精度,提出一种蚁群小波神经网络伪点迹估计器来事先重构UUV的运行轨迹;小波神经网络一方面具有很高的逼近精度,另一方面具有很快的收敛速度,同时它还拥有快速脱离局部极小点的特性,比较适用于多传感器数据融合领域;因此,文中采用小波神经网络构造位置估计器进行伪点迹重构;然而由于神经网络规模不断增加导致收敛速度大大降低,为解决这一矛盾,采用一种基于空间网格机理的蚁群算法来优化小波神经网络伪点迹估计器,从而形成蚁群小波神经网络伪点迹估计器;仿真实验中共有3台UUV参与定位,一共选取1 500组来自于传感器的数据,前1 000组用于估计器的优化训练,后500组用于测试训练后伪点迹估计器的性能。蚁群算法中蚂蚁规模为50,经验因子为α=0.5,启发因子为β=0.3,挥发系数初值为ρ（0）=1.0,信息素初值为τij（0）=constant,信息素总量取Q=1,解空间分区数n1=n2=n3=n4=n5=6,m={3,6,9,3};结果表明,蚁群小波神经网络伪点迹估计器UUV互定位误差小于遗传小波神经网络伪点迹估计器UUV和传统定位模式UUV,其构成的多UUV系统具有更高的互定位精度。

英文摘要：

In order to improve mutual positioning precision for MUUVS under weak communication conditions, a kind of pseudo point trajectory estimator is proposed to reconstruct UUV trajectory in this paper beforehand. Wavelet neural network （WNN） has not only higher approximate precision but also more rapid convergence rate. The WNN can escape from local minimize value quickly. It is adaptive for multi --sensor datum fusion system. So we reconstruct pseudo trajectory by position estimator based on WNN. However, with more and more large dimensions of WNN, its convergence rate will decrease greatly. To solve the contradiction, we will adjust the WNN pseudo trajectory estimator by an enhanced ant colony algorithm based on space grid principle; so we acquire an ant colony WNN pseudo trajectory estimator. Three UUVs will be required to participate in mutual positioning in the simulation experiment; We will obtain 1500 groups datum from rele- vant sensors, the former 1000 groups is used for training the estimator, the others will testify the performance for trained pseudo measure- ment estimation. The ant scale of ant colony algorithm is 50 , the experience factor isa = 0.5 , the enlightenment factor isfl= 0.3 , the ini- tial value of pheromone is ri~ （0） = const , the initial value of Volatilization Coefficient isp（0） = 1.0 , the initial value for the amount of pher- omone is Q = 1 , the partition number of solution space is nl = nz = rt3 = n4 = n~ = 6 , rn = { 3,6,9,3} . The results demonstrate that the wavelet neural network pseudo measurement estimator adjusted by ant colony algorithm for UUV is less mutual positioning error than the genetic wavelet neural network pseudo trajectory estimator and traditional mutual position methods on the whole. Furthermore, Simulation results demonstrate that the estimator will obtain higher mutual positioning precision.