Research On Visible Light Indoor Localization Algorithm Based On Elman Neural Network

Since Global Positioning System(GPS)and other Global Navigation Satellite Systems(GNSS)are unable to provide accurate indoor positioning information services,research on indoor positioning technology has become the focus of scholars’ attention.Compared with traditional indoor positioning technology,visible light indoor positioning technology has the advantages of easy construction,low positioning cost,high positioning accuracy,high security,and no RF or electromagnetic interference.Therefore,this thesis takes visible light indoor positioning technology as the research direction,and the main work includes:(1)This thesis analyses commonly used visible light indoor positioning algorithms,including image sensor-based,photodetector(PD)-based and machine learning-based visible light indoor positioning algorithms.Through comparison and analysis with the commonly used visible indoor positioning algorithms,this thesis decides to take the machine learning-based visible indoor positioning algorithm as the research direction to solve the problems of low positioning accuracy and poor robustness of traditional visible indoor positioning algorithms.(2)This thesis establishes a visible light indoor localization system consisting of a single LED lamp and multiple PDs,with the LED lamp placed at the top centre of the room as the transmitter,four horizontal PDs placed at the four corners of the square model as the receiver,and the location to be measured located at the centre of the square model.The Elman neural network is introduced into the visible light indoor positioning algorithm,and two Elman neural networks are used to predict the horizontal and vertical coordinates of the points to be measured to achieve the initial positioning of the points to be measured.Aiming at the problem of large positioning errors of individual points in the positioning results of the Elman neural network,Weighted Knearest Neighbor(WKNN)algorithm is introduced to further modify the location results based on Elman neural network.The simulation results show that the algorithm in this thesis can obtain good localisation results.The average localisation error is 4.75 cm in an indoor environment of2.25×2.25×2.6m,which is 54.55% better than the classical BP neural network and 41.57% better than the single Elman neural network.The average positioning time of this algorithm is 0.197432 s,which is within 1s and can meet the needs of production and life in most cases.(3)An experimental environment was set up to test the above visible light indoor positioning system consisting of a single LED lamp with multiple PDs.The test results showed that the average positioning error of this algorithm was 10.46 cm and the average positioning time was0.200429 s.Compared with the BP neural network in the same experimental environment,the positioning accuracy of this algorithm was improved by 49.81%.Compared with a single Elman neural network in the same experimental environment,the localization accuracy of this algorithm improved by 43.24%.The test results in the experimental environment show that the algorithm of this thesis is effective.