Multi-sensor Data Co-location Technology Based On Machine Learning

Indoor positioning is needed in hospitals,warehouses,shopping malls,tunnels and other places.The research content of this article is based on machine learning,a variety of sensor data collaborative positioning technology,mainly in conjunction with Wi-Fi and inertial sensors for positioning.The main work and innovations of this article are as follows:(1)Aiming at the defects of the existing channel status information Channel Status Information(CSI)image positioning,a Wi-Fi positioning method based on a new type of CSI image and convolutional neural network is proposed.First,theoretically verify the feasibility of using Ao A-To F(Angle of Arrival-Time of Flight)pseudo-spectrum(an image containing the angle of arrival and flight time of wireless signals)for indoor positioning.Among them,a soft phase error cancellation is proposed The method eliminates the channel error of CSI.Then use the Ao A-To F pseudo-spectrum as the input of the Deep Convolution Neural Network(DCNN).The algorithm automatically extracts the data features of the Ao A-To F pseudo-spectrum by performing four convolution and pooling layers,using the BP algorithm Get training weights.Finally,an improved probabilistic positioning method is proposed for positioning.(2)Aiming at the problems of existing multi-sensors,a multi-sensor co-location technology based on dual neural networks is proposed.First introduce the architecture of the collaborative positioning model,and then focus on the trajectory model of inertial sensors.The finally established multi-sensor joint indoor positioning model is to introduce inertial sensor data information on the basis of the Wi-Fi positioning model,and use the Gated Recurrent Unit(GRU)to establish the motion trajectory model,which improves the accuracy and anti-interference ability of indoor positioning,making it complicated.In the indoor environment,it can still maintain good positioning accuracy.(3)Test the performance of the Wi-Fi positioning method and co-location technology proposed in this paper on the basis of the measured data set.The experimental results show that the Wi-Fi positioning method based on the new CSI image and convolutional neural network is superior to the positioning method using CSI original data positioning or the original data to construct the CSI image;the multi-sensor cooperative positioning technology based on dual neural network achieves inertia The advantages of the sensor and Wi-Fi sensor complement each other,which further improves the positioning accuracy and anti-interference ability.