Research On Indoor 3D Visible Light Positioning System Based On Intelligent Algorithm

The maturity and field deployment application of 5G communication technology is a great leap in the field of communication,but the near-saturated spectrum utilization has become a major resistance to limit the development of radio frequency communication.Visible light communication(VLC),with its advantages of unlimited bandwidth and no need for spectrum authentication,effectively makes up for the lack of spectrum resources for radio frequency communication,and is regarded as a research direction with great development potential in the new generation of wireless communication technology.It is expected to become a key technology for 6G communication.With the development of VLC technology and the popularity of LED lights,VLC-based visible light positioning(VLP)technology has become a research hotspot,with the advantages of anti-electromagnetic interference,low cost,high speed,safety,energy saving,green and environmental protection,etc.However,the current research on VLP is mostly focused on two-dimensional positioning,and less research is conducted for high-precision three-dimensional positioning.Due to the powerful nonlinear system processing capability of intelligent algorithms such as smart swarm perception algorithm and neural network,the research of VLP based on intelligent algorithms is in full swing.In this paper,we focus on the indoor visible light positioning system based on intelligent algorithms,and the main research contents are as follows.First of all,the initial weights and thresholds of the extreme learning machine(ELM)neural network are determined by using the sparrow search algorithm(ISSA)optimized by circle chaotic mapping.An indoor two-dimensional visible light positioning method based on ISSA-ELM neural network is proposed to establish a position prediction model considering the reflection effects of ceiling,wall and floor.The results show that the training time of the proposed method is 0.1920s,the positioning time is 1.0ms,and the average positioning error in the height plane of the 0m-1.5m range is less than 5cm in a room of 5m×5m×3m.The comparative analysis of this method with various other classical positioning methods further verifies that this method has significant advantages in positioning accuracy,training speed and positioning time.Secondly,by increasing the number of LED light sources for light source layout,the problem of uneven distribution of received optical power in the higher receiving plane of the room is solved,and then the indoor 3D visible light positioning method based on ISSA-ELM neural network is proposed and the prediction model is established.The results show that the training time of the proposed method is 3.20min,the positioning time is 0.1478s,and the average positioning error is less than 5.5 cm and the maximum positioning error is less than 15 cm in a room of 5m×5m×3m covering all the areas of 5m×5m×2.9m.The method is compared with the indoor 3D visible light positioning methods based on BP,ELM,SSA-ELM neural networks,which shows that this method has comprehensive positioning range coverage,high positioning accuracy.fast training speed and short positioning delay.Thirdly,the optical radiation characteristics of Lambert light source and three non-Lambert light sources,including EdiPower light source,LUXEON Rebel light source and Z-Power light source are simulated and analyzed.The indoor 3D visible light positioning method based on non-Lambert light source is studied and compared with the positioning system based on Lambert light source.The results show that the average positioning error of the system based on non-Lambert light sources is slightly higher than that based on Lambert light source,but it can still meet the demand for 3D positioning accuracy.When the positioning height is lower than 2.5m,the three non-Lambert light sources are suitable for indoor 3D visible positioning.If the positioning height is higher than 2.5m,the EdiPower light source and LUXEON Rebel light source no longer meet the requirements,and Z-Power light source is recommended.Finally,three sparse optimization schemes of training data set,such as grid sparse method,concentric circle sparse method and snowflake sparse method,are proposed to realize indoor 3D visible light positioning based on sparse training set.The sparsification degree of the training set is greater than 50%for all three schemes,which reduces the training time of the ISSA-ELM neural network by more than 60%.In addition,the average positioning error of the system is less than 2cm and the maximum positioning error is less than 20cm after optimizing the training set by the three sparsification schemes,which ensure the high positioning accuracy of the positioning system while effectively sparsely optimize the training set and reduce the size of the training set.