基于BPNN-MLR的室内可见光定位算法

Indoor Visible Light Positioning Algorithm based on BPNN-MLR

针对传统位置指纹算法存在定位精度低和计算复杂度高等问题,文章提出了一种基于反向传播神经网络(BPNN)和多元线性回归(MLR)的单发光二极管(LED)室内定位算法。首先,利用3个水平光电探测器(PD)作为接收器接收光功率,待测点位于接收器的中心;然后根据接收到的光功率向量,利用BPNN确定待测点粗略的位置范围;最后以该位置范围作为约束条件,利用MLR对待测点的位置进行更精确地定位。实验结果表明,在2.0 m×2.0 m×2.5 m的室内空间中,该算法的平均定位误差为5.04 cm,平均定位时间为0.00283 s。与传统的位置指纹算法相比,该算法的平均定位精度提高了41.53%,平均定位时间减少了56.60%,在较低计算复杂度的前提下实现了更精确的定位。

Aiming at the problems of low positioning accuracy and high computational complexity of the traditional location fingerprint algorithm,this paper proposes a single Light Emitting Diode(LED)indoor positioning algorithm based on Back Propagation Neural Network and Multiple Linear Regression(MLR).First,three horizontal PhotoDetectors(PD)are used at the receiver side to receive optical power,and the point to be measured is located at the center of the receivers.Then,according to the received optical power vector,the BP neural network is used to obtain the coarse location range of the point to be measured.Finally,the location range is used as a constraint,and multiple linear regression is used to achieve more accurate positioning of the point.The experimental results demonstrate that in an indoor space of 2.0 m×2.0 m×2.5 m,the average positioning error of the algorithm is 5.04 cm,and the average positioning time is 0.00283 s.Compared with the traditional location fingerprint algorithm,the average positioning accuracy of the algorithm is improved by 41.53%,and the average positioning time is reduced by 56.60%,which achieves more accurate positioning under the requirement of lower computational complexity.