Study On An Autonomous Indoor Localization Technology Using Millimeter Wave Radar

With the rapid development of Internet technology,people’s demand for Location Based Services（LBS）is increasing,and ubiquitous autonomous indoor positioning has become an emerging demand.Compared with outdoor positioning,indoor positioning requires higher accuracy and is prone to interference by multipath propagation.Traditional indoor positioning methods usually rely on Wi Fi,Bluetooth,Visible light and other beacons.However,these indoor positioning methods need additional laying and regular maintenance of beacons,which do not meet the university requirements of autonomous positioning.Therefore,high-precision autonomous positioning method is becoming more concerned.In recent years,the radar-based indoor positioning method is a hot spot in the research of the indoor autonomous positioning method.With the arrival of the 5^thGeneration Mobile Communication Technology（5G）,millimeter wave radar has attracted more and more attention,due to its high accuracy,low cost and low power consumption.In this work,an indoor pedestrian autonomous positioning algorithm based on millimeter wave radar is innovatively proposed.The algorithm mainly uses millimeter wave radar and Inertial Measurement Unit（IMU）inside the mobile phone to independently locate pedestrians without laying and using beacons in the environment.The algorithm first uses IMU data to generate the original trajectory of pedestrians,then uses the position and intensity information of point cloud collected by millimeter wave radar for loop closure detection,and finally optimizes the trajectory through the back-end of Graph SLAM algorithm.The main contributions are as follows:Firstly,this paper proposes a feature extraction method based on improved point cloud Auto-Encoder to solve the disorder,unstructured and sparse of millimeter wave radar point cloud.Due to the disorder and unstructured features of millimeter wave radar point clouds,it will bring difficulties to the similarity calculation in loop closure detection,and the sparsity of millimeter wave radar point cloud will also affect the accuracy of loop closure detection.This paper proposes to use Point Net Auto-Encoder neural network to process millimeter wave radar point cloud,extract more structured features from sparse and unstructured point cloud,and calculate the similarity of point cloud position information.Secondly,This paper proposes a loop closure detection method based on improved multi-feature information fusion.Because millimeter wave radar is easy to the interference of multipath propagation,this will result in a decrease in the accuracy of loop closure detection.The loop closure detection of multi-feature information fusion can combine the intensity information and position information of point cloud,and further proposes to remove the over-matched loop closure to increase the accuracy and robustness of the loop closure detection.Finally,This paper proposes a back-end method based on an improved Graph SLAM algorithm.The method can allocate weights according to the similarity of different loop closures,so as to enhance the accuracy and robustness of the back-end.Moreover,This paper provides a comprehensive set of experimental results to evaluate the performance of the proposed algorithm.The results show that the absolute trajectory error of the autonomous pedestrian indoor positioning algorithm can reach 0.5m without additional beacons,and the relative pose error is 0.7m.