基于改进ELM的煤矿井下定位算法

Coal mine underground positioning algorithm based on improved elm

摘要:

提出了一种基于改进极限学习机（Extreme Learning Machine ELM）神经网络的煤矿井下人员定位算法，针对测距模型易受井下复杂环境干扰，无法准确测距的问题，选用基于指纹的位置匹配模型。使用极限学习机将指纹和位置进行匹配，选用改进鲸鱼优化算法（Improved Whale Optimization Algorithm IWOA）选取ELM合适的输入权值和隐含层阈值，以提高定位精度。在定位的在线阶段，将新的指纹数据代入带动态权值因子的在线顺序极限学习机（Dynamic Weight Factor Online Sequential Extreme Learning Machine DOS-ELM）模型对定位模型进行动态调整，以克服电磁传播环境变动使定位结果产生的误差。仿真实验结果表明，该模型的定位误差在1.5m以内的置信概率为72%，平均定位误差为1.64m，与其他算法的实验结果相比，本文算法鲁棒性强，定位精度高

Abstract:

In this paper, an adaptive location algorithm based on the improved extreme learning machine (ELM) neural network is proposed. Aiming at the problem that the ranging model is easily disturbed by the complex underground environment and the ranging error is large, the fingerprint based location matching model is selected. ELM is used to match fingerprint and location. Improved whale optimization algorithm (IWOA) is used to select elm's appropriate input weight and hidden layer threshold to improve positioning accuracy. In the online stage of localization, the new fingerprint data is substituted into the dynamic weight factor online sequential extreme learning machine (DOS-ELM) model with dynamic weight factor to dynamically adjust the localization model, so as to overcome the error caused by the change of electromagnetic propagation environment. The simulation results show that the confidence probability of the positioning error within 1.5m is 72%, and the average positioning error is 1.69m. Compared with the experimental results of other algorithms, this algorithm has strong robustness and high positioning accuracy.