基于遗传算法优化BP神经网络的可见光定位

A visible light localization algorithm based on BP neural network optimized by genetic algorithm

采用接收信号强度(RSS)方法的室内可见光定位,因受多径效应及噪声的影响,对距离估计不准确,定位精度不高。为提高定位精度,本文提出了一种采用遗传算法优化BP神经网络(GA-BP)的距离估计方法。先通过遗传算法优化BP神经网络的初始权值,经过优化后的BP神经网络收敛速度快,不易限于局部最优。再利用GA-BP神经网络对收发端之间的距离进行修正,使其接近于真实距离。最后使用最小二乘法解算待定位点坐标,同时在不同定位范围和不同定位位置下,与传统RSS加权质心方法的可见光定位结果进行对比。仿真结果表明,在5m×5m×3m的定位场景中,平均定位误差可以达到0.6428cm。与传统RSS加权质心方法相比,平均定位精度提高了约96.4%。且在不同定位范围和不同定位位置下,平均定位误差稳定在毫米级,尤其不随定位范围的扩大而扩大。有效地提高了室内定位精度和系统应用的普适性。

Due to multipath effects and noise,the indoor visible light positioning which adopts the received signal strength(RSS) method is inaccurate in estimating the distance,resulting in low positioning accuracy.This paper proposes a distance estimation method based on BP neural network optimized by genetic algorithm,which improves the positioning accuracy.Firstly,the initial weight of BP neural network is optimized by genetic algorithm.The optimized BP neural network has fast convergence speed and is not easy to be limited to local optimum.The GA-BP neural network is then used to correct the distance between the transceiver terminals,so that the corrected distance is close to the real distance.Finally,the least squares method is used to solve the coordinates of the anchor points.At the same time,the GA-BP method is compared with the traditional RSS weighted centroid method under different positioning ranges and different positioning locations.The simulation results show that the average positioning error can reach 0.642 8 cm in the 5 m×5 m×3 m positioning area.Compared with the traditional RSS weighted centroid method,the average positioning accuracy improves by about 96.4%.Within different positioning ranges and locations,the average positioning error is stable at the millimeter level and will not expand with the expansion of the positioning area,which effectively improves indoor positioning accuracy and universal applicability of system applications.