After analyzing the existing ambiguity resolution methods the dissertation investigates the baseline rotation method based on the antenna - swap idea , and develops the fixed axis rotation maximum - minimum method ? farmm and the equivalent rotation maximum - minimum method ? ermm applicable for static basement and moving vehicle respectively . the former lowers the demands on rotation device while the latter improves the flexibility of antennae installation . meanwhile the dissertation also brings forward an ambiguity function search method which needs only one single epoch data with the outer aided information , and investigates the accuracy requirements for the assistant pointing information 其次,通过与gps等多星系统定向关键技术的对比分析,指出双星定向模糊度确定的难度;在分析现有模糊度求解方法的基础之上,研究了基于交换天线思想的转动基线模糊度确定方法,并提出了适合于静基座的定轴转动极大/极小值farmm法,以及适合于地面运动载体的等效转动极大/极小值ermm法,前者降低了对转动机构的要求,后者则提高了天线安装的灵活性;同时,提出了一种借助外部辅助信息的模糊度函数搜索法实现单历元双星定向,并分析了对外部辅助单轴姿态信息的精度要求。
The quasi - physical method makes the original problem an optimization problem in mathematics . there is often the possibility of going to a local minimum of object function when we solve the optimization problem mathematically . as for how to jump out of the trap of local minimum so that the calculation can head for a region with better prospects , the quasi - physical method is helpless 拟物方法将原始问题落实为优化问题，而用数学方法在求解优化问题时，常常会碰到计算落入目标函数的局部极小值陷阶的困境，如何从这种困境中逃逸出来，使得计算奔向前景更好的区域，拟物方法则无能为力，而应用拟人方法则可以设计出好的“跳出陷阱”策略。
However , the neural network easily falls into local minimum , and weakly search the overall situation . the genetic algorithm ( ga ) has the ability of searching overall situation . the genetic neural network recombines the genetic algorithm ’ s of seeking the superior overall situation and the neural network ’ s nonlinear characteristic and rapid convergence 但神经网络具有易陷入局部极小值以及全局搜索能力弱等缺点;而遗传算法具有较好的全局最优搜索能力,遗传神经网络将两者结合,既保留了遗传算法的全局寻优的特点,又兼有神经网络的非线性特性和收敛的快速性。
Neural network system which can self study and self adapt can well solve the shortcoming , there still have the shortcoming of low training speed ， apt to get into local minimal value and poor macrocosm searching ability for neural network . after studying genetic algorithms , it was found that ga can better mend the shortcoming 但是神经网络存在训练速度慢，易陷入局域极小值和全局搜索能力弱等缺点。在研究了遗传算法后，发现它可以较好地改进以上的缺点。所以在此基础上，利用进化神经网络，采取bp算法和遗传算法建立了真核生物基因启动子识别模型。
