CAJ | 학술논문

煤矿井下受限空间中，射频信号强度受到多径衰落、阴影效应及人为因素的影响，采用路径损耗模型的定位方法误差较大，提出了基于核函数法及粒子滤波的定位算法。该算法利用指纹匹配技术结合贝叶斯估计，基于核函数法构建模型，搜索训练数据中接近未知节点指纹特征的位置并加权得到初步观测坐标，最后利用粒子滤波将目标运动状态与观测值相融合，平滑位置突变以追踪移动轨迹。实验证明，对于静态目标定位，核函数法效果优于确定型匹配算法和高斯分布模型；对于动态目标定位，所提算法比基于Markov状态转移的算法定位结果更精准。
매광정하수한공간중，사빈신호강도수도다경쇠락、음영효응급인위인소적영향，채용로경손모모형적정위방법오차교대，제출료기우핵함수법급입자려파적정위산법。해산법이용지문필배기술결합패협사고계，기우핵함수법구건모형，수색훈련수거중접근미지절점지문특정적위치병가권득도초보관측좌표，최후이용입자려파장목표운동상태여관측치상융합，평활위치돌변이추종이동궤적。실험증명，대우정태목표정위，핵함수법효과우우학정형필배산법화고사분포모형；대우동태목표정위，소제산법비기우Markov상태전이적산법정위결과경정준。
In the confined coal mine environment,RF signal strength is affected by the multipath fading,shadow effect and human factors. Utilizing a path loss model for position makes the error larger,so a novel positioning algorithm is proposed based on kernel function and the particle filter. By constructing a model based on kernel function, the algorithm combines the fingerprint matching technology with Bayesian estimation to search for training data,which are closer to unknown-node fingerprint characteristics. These locations of training data are weighted as preliminary observations. Finally, using particle filter integrates target motion state with observations for smoothing position mutations and tracking trajectories. Experiments show that, for static target location, the kernel function method outperforms the algorithms with deterministic matching or Gaussian distribution model;for dynamic targeting, positioning results of the proposed algorithm are more accurate than an algorithm using Markov state transition.