Research On WIFI And Multi-sensor Fusion Positioning Algorithm Based On Machine Learning

With the advancement of urbanization,the scale of buildings has become larger and larger,and the demand for location is not only satisfied with the outdoor.High-precision indoor positioning can enhance the industrial value of catering and logistics.However,the indoor environment is complex and changeable.The GNSS signal cannot be used for indoor positioning due to the occlusion of walls,etc.Therefore,it is a difficult problem to obtain an indoor positioning technology with both stability and high precision.In recent years,with the continuous development of indoor positioning research,positioning data sources have become more diversified.Facing the inevitable shortcomings of a single positioning technology,scholars have begun to integrate different positioning technologies to complement each other to ensure positioning accuracy while improving positioning stability.In view of the abovementioned problems in indoor positioning,this paper has done the following research work:(1)Aiming at the problems of large positioning fluctuation caused by the redundant information of the fingerprint database matching in indoor positioning,and too many samples in the database,the positioning timeliness is poor,and an indoor positioning method based on the firefly algorithm optimized support vector machine is proposed.Use singular spectrum analysis to preprocess the data to remove noise,optimize the parameters of the support vector machine through the firefly algorithm,and establish an indoor positioning regression model.Experimental results show that the average positioning error of the algorithm in this paper is 24.2%lower than that of PSO-SVM algorithm,and 33.3%lower than that of GA-SVM algorithm;the minimum positioning error is reduced by 75%compared with PSO-SVM algorithm and 66.7%compared with GA-SVM algorithm.In addition,the maximum positioning error is 46.3%lower than the PSO-SVM algorithm,and 62.9%lower than the GASVM algorithm.Compared with the current indoor positioning methods,the algorithm in this paper has a fast convergence speed,which improves the accuracy and stability of indoor positioning.(2)For the estimation of the step length difference of different pedestrians in the pedestrian dead reckoning,a nonlinear step length model is used to estimate the step length,and the model is improved,and the walking frequency and acceleration variance are introduced to the step length.For estimation,the step length can be adjusted adaptively with the change of step frequency.The experimental results show that the average positioning error of the algorithm in this paper is 1.42m,which is about 20%lower than that of the traditional PDR algorithm,which effectively improves the positioning performance.(3)Aiming at the problem of low accuracy of single positioning technology,this paper introduces chaotic variables and seagull algorithm to optimize the BP neural network model,improves the accuracy of data fusion results through adaptive hidden layer neurons,and uses the seagull intelligent optimization algorithm to speed up convergence The speed obtains the global optimal solution,and the chaotic variable is introduced to initialize the seagull position to solve the problem that the optimization algorithm is easy to fall into the local optimal solution.The experimental results show that the average positioning error of the chaotic variable optimization algorithm is less than 1m and the positioning path fits the actual route.It has high robustness and can better integrate the WIFI positioning results and the multi-sensor positioning results.(4)In view of the path planning RRT*algorithm search path is too long,the convergence speed is too slow,this paper draws on the advantages of the RRT-connect algorithm to grow two fast-expanding random tree search map space from the starting point and the target point at the same time,and optimizes the traditional RRT*algorithm.Experimental results show that the algorithm in this paper reduces the number of iterations by 50%and the planning time by 50%compared with the RRT*algorithm.At the same time,the relationship between air pressure and altitude is used to solve the problem of floor positioning.Figure[53]Table[15]Reference[81]...