基于粒子滤波的定位系统中累计误差消除的进化策略

Evolutionary strategy for elimination of accumulated errors in positioning system based on particle filter

粒子滤波算法是地磁定位过程中采用的一种常用方法,但是该方法有一个致命缺点,即存在累积误差,会导致定位的失败。文中引入进化算法思想改进了粒子滤波算法,通过计算适应度值来控制变异步长,自适应地提高进化策略的搜索效率和精度,且能有效地提升重采样之后粒子的丰富性,然后依据粒子的权重实现优选。在移动一段距离之后,周期性地进行轨迹地磁匹配运算,克服累积误差对当前时刻的影响。同时在地磁匹配的过程中,在精确匹配之前采用一个预匹配过程,大大减少匹配时间。通过C++仿真及实地定位测试试验,结果表明该方法能够有效地提高粒子滤波性能及定位精度。

The particle filter algorithm is a commonly used method for a localization based on magnetic measurement, however, it has a fatal flaw, called the existence of accumulated errors, leading to the failure of localization. According to the mutation step controlled by fitness, an adaptive evolution strategy is proposed in the particle filter algorithm to improve the searching efficiency and the precision, thus enhanceing the variety of resampled particles. Then, the optimization of selecting particles is realized based on the particle weight. To increase the positioning accuracy and overcome the effects on accumulated errors, a geomagnetic matching algorithm is periodically called after the target moving some steps. In the geomagnetic matching, the use of pre-matching prior to exactly matching process can reduce the convergence time. The simulation by C ＋ ＋ on an Android smartphone, the test in an indoor environment and further simulation based on real-world measurements show that the algorithm can effectively improve the filter performance and the positioning accuracy.