Indoor Visible Light Positioning Based on Backward Propagation Neural Network and Improved Osprey Optimization Algorithm

Visible Light Positioning (VLP) has been widely recognized as a promising indoor positioning solution. As one of the two typical VLP technologies, range-based VLP has gained significant attention due to its low cost and easy to implement. In most existing range-based VLP methods, the received signal strength (RSS) from each light-emitting diode (LED) lamp needs to be converted into distance to estimate the target position. However, due to the effect of multipath and noise, the error of RSS conversion to distance cannot be ignored, which usually leads to a large positioning error. In this paper, an improved range-based VLP method is presented to address the above problem. In the proposed method, the RSS is firstly fed into a backward propagation (BP) neural network for distance correction to improve the ranging accuracy. Then, the Least Squares method (LSM) is utilized to obtain the positioning area affiliated with the target node (TN), and the positioning problem is transformed into an optimization problem, which can be solved by using intelligent optimization algorithms, such as the osprey optimization algorithm (OOA). In order to achieve higher positioning accuracy, an improved OOA (IOOA) algorithm is designed in this paper, in which strategies such as Sobol sequence distribution and Brownian motion are employed to address the issues of slow convergence and susceptibility to local optima in the original OOA algorithm. Simulation results show that, when the signal-to-noise ratio (SNR) is $\mathbf{1 5} \mathbf{~ d B}$, the ranging error can decrease by 26.3 % after the BP neural network is used. Compared to the original OOA algorithm, the positioning accuracy of the BP-OOA and BP-IOOA algorithms can be improved by 40.53 % and 66 %, respectively.