基于粒子群优化神经网络的列检人员位置融合导航定位方法

Position Fusion Navigation and Location Method of Train Inspectors Based on Particle Swarm Optimization Neural Network

列检场地较大,人员是否在安全区域内难以判断,为了实时、合理、精确地获得列检人员的位置信息,提出基于粒子群优化神经网络的列检人员位置融合导航定位方法。确定列检安全作业综合管理平台层级架构,将该平台作为导航算法实现基础,以惯性定向定位导航和全球导航卫星定位系统结合的方式,采集列检人员的初始位置、运动速度等信息,保证基本输入输出过程的针对性。分析粒子群优化过程,包括粒子初始化和种群评估,确定网络架构,选择激活函数,建立经典神经网络模型,保证融合导航过程的合理性。通过粒子群优化神经网络的方法提高网络搜索能力,避免陷入局部最优。将采集到的融合信息作为网络输入,设置连接权值和网络相关参数,根据适应度值,更新粒子速度与位置。当收敛精度满足要求时,输出导航定位结果。实验结果表明,该方法定位结果和实际位置的吻合度较高;仅需300次迭代即可实现算法收敛,训练误差基本保持在0.002以下。可以有效定位到列检人员的作业具体位置,提高了列检人员导航定位实时性。

The train inspection site is large,and it is difficult to judge whether a personn is in the safe area.In order to obtain the position information of the train inspector in real time,reasonably and accurately,a train inspector position fusion navigation positioning method based on particle swarm optimization neural network is proposed.This paper determines the hierarchical structure of the comprehensive management platform for train inspection safety operation,and uses the platform as the basis for navigation algorithm implementation.It collects the initial position and movement speed of train inspectors in the way of combining inertial orientation navigation and global navigation satellite positioning system to ensure the pertinence of the basic input and output process.It analyzes the process of particle swarm optimization,including particle initialization and population evaluation,determines the network architecture,select the activation function,and establishes the classical neural network model to ensure the rationality of the fusion navigation process.The method of particle swarm optimization(PSO)is used to improve the network searching ability and avoid falling into local optimum.The collected fusion information is taken as the network input,the connection weight values and network related parameters are set,and the particle speed and position are updated according to the fitness value.When the convergence accuracy meets the requirements,the navigation and positioning results are output.The experimental results show that the positioning results of this method agree well with the actual position.Only 300 iterations are needed to achieve the algorithm convergence,and the training error is basically kept below 0.002.It can effectively locate the specific operation position of the train inspector,improving the real-time navigation and positioning of the train inspector.