Research On Improved GRNN Indoor Positioning Based On HBA-BPNN Calibration Model

With the rapid growth of the number of mobile users around the world,and users are mostly in indoor environments,the demand for effective indoor positioning services is growing.However,as the development of large-scale integrated circuit,intelligent devices are being updated faster and faster,and their built-in software and hardware configurations are also different.The difference of software and hardware configuration between different intelligent devices will lead to the inconsistency of Bluetooth RSSI signals received by them,which will affect the indoor positioning accuracy,and ultimately lead to the unavailability of traditional fingerprint positioning methods,which cannot meet people’s positioning needs.To solve theproblems mentioned above,this paper proposes an improved GRNN(Generalized Regression Neural Network)indoor fingerprint location method based on HBA(Honey Badger Algorithm)optimized BP neural network calibration model.The work of this paper mainly contains the following respects:(1)Firstly,this paper proposes an indoor Bluetooth RSSI calibration algorithm based on Honey Badger Algorithm Back Propagation Neural Network(HBA-BPNN)model to calibrate the received Signal Strength Indicator(RSSI)received by Bluetooth sensors of different intelligent mobile terminal devices,making the Bluetooth RSSI signal strength received by different intelligent mobile terminal devices at the same location tends to be consistent,so as to solve the problem of heterogeneous software and hardware among different intelligent mobile terminal devices.The calibration algorithm uses the searching speed and global searching ability of the Honey Badger algorithm to help the BPNN model select the optimal initial weight and threshold,effectively avoiding the disadvantage that the BPNN model is easy to fall into the local optimum.In order to obtain the best calibration effect for the HBA-BPNN calibration algorithm,we explored and analyzed the influence of the different parameter β,C value on the calibration effect of HBA-BPNN calibration algorithm to select the best β,C parameter values are combined for subsequent calibration and positioning experiments.(2)Secondly,to solve the problem that using traditional RSSI fingerprint location model for indoor location will lead to large location error,this manuscript proposes an indoor fingerprint positioning algorithm based on a kind of improved GRNN(Generalized Regression Neural Network)model.This algorithm uses ACO(Ant Colony Optimization)to optimize the smoothing factor of GRNN to get a better indoor positioning model.In order to ensure that the indoor fingerprint location algorithm based on ACO-GRNN model finally obtains the optimal location result,we compared the effects of different training sets and different iterations of the ACO algorithm on the location accuracy,so as to select the optimal training set and iterations for subsequent experiments.(3)Finally,the two algorithms are effectively combined and compared with different calibration algorithms and different positioning algorithms.The experimental results show that compared with the uncalibrated data set and the data set calibrated by the calibration algorithm based on BPNN model and WOA BPNN model,the average positioning error of the data set calibrated by the test intelligent mobile terminal equipment using the calibration algorithm based on HBA-BPNN model proposed in this paper is reduced by 65.3%,45.8% and 23.6% respectively,It effectively eliminates the indoor positioning error caused by heterogeneous software and hardware of different intelligent mobile terminal devices;Compared with the indoor fingerprint location algorithm based on BPNN model and traditional GRNN model,the average location error of the intelligent mobile terminal equipment tested using the indoor fingerprint location algorithm based on the improved GRNN model proposed in this paper is reduced by 53.6% and 30% respectively,and the indoor fingerprint location algorithm based on ACO-GRNN model can achieve sub meter level location accuracy.In conclusion,by effectively combining the two algorithms proposed in this paper,we can effectively solve the problem of software and hardware heterogeneity between different intelligent mobile terminal devices,and improve the positioning accuracy.