Research Of Indoor Positioning Technology And Trajectory Prediction Based On Machine Learning

With the prosperity of IOT,people put forward higher and higher requirements for indoor positioning and trajectory prediction services.Due to the influence of sheltering,GPS technology can't meet the needs of people's indoor positioning.Therefore,indoor positioning technology is developing vigorously,and it has been a hot research and application field since its birth.This paper adopts WIFI indoor positioning technology.The indoor location and trajectory prediction are realized by using the fingerprint method.Aiming at the problem of WIFI router deployment,the AP comprehensive selection strategy is proposed.The strategy selects the frequency of AP signal and the variance of signal strength received by RP point.It makes positioning error minimum and stability maximum.To solve the problem of high computation cost,an R-FCM algorithm is proposed.This method makes the boundary elements be effectively distinguished.It reduces the calculation cost and improves the positioning accuracy simultaneously.Aiming at the problem of low positioning accuracy,the R-KNN algorithm is proposed,which improves the average accuracy significantly.Based on transfer learning,a method of transplanting fingerprint database is proposed.This method greatly reduces the computation cost.Aiming at the problem of extraction of stay points,a method of distance and speed limitation is proposed.Compared with the traditional Markov model,the accuracy of LSTM prediction model is improved to some extent,and the accuracy is excellent.