The Research On The Multi-sensor-based Indoor Robot Integrated Navigation Method

With the rapid development of robot technology,robots gradually play an important role in our daily life.As the key premise for mobile robots to complete various tasks,the demand for the accuracy of robot positioning is also increasing.In the existing Global Navigation Satellite System(GNSS),mainly Global Positioning System(GPS)and BeiDou Navigation Satellite System(BDS),it can achieve the positioning of the robot.However,under indoor conditions,GNSS cannot meet the mobile robot's needs for indoor high-precision positioning and navigation due to signal occlusion,and indoor positioning technology has gradually become a research hotspot.For indoor mobile robot combination positioning,Firstly,this article introduces several commonly used indoor positioning technologies,and analyzes the advantages and disadvantages of the current indoor positioning technology.It can be seen from the analysis that in the existing positioning technology,the positioning error of Dead Reckoning(DR)accumulates with time,and the positioning error of Light Detection And Ranging(LiDAR)is independent at each moment and does not accumulate with time.To take advantage of the two sensors to overcome their deficiencies,this paper presents the DR/LiDAR integrated positioning system for indoor robot.Then we built a DR/LiDAR integrated positioning experiment platform,and developed data acquisition software,it laid the foundation for the verification of the algorithm.Based on DR/LiDAR integrated positioning model,this paper studies the combination of loose and tight positioning,the loose and tight combination positioning model of DR/LiDAR is constructed,and the performance of the model is analyzed through experiments.The experimental results show that the integrated combination method has higher positioning accuracy and smaller error fluctuation.On this basis,the traditional data fusion algorithm Extended Kalman Filter(EKF)has high positioning accuracy and fast convergence speed,however,the prediction and estimation accuracy of EKF depends on the accuracy of noise statistical characteristics.Once the difference between the noise description and the actual situation is too large,EKF positioning performance will decline rapidly and even diverge.In practical application,the true values of process noise variance matrix Q and measurement noise variance matrix R are not easy to determine.In order to improve the robustness of data fusion algorithm,this paper combines Particle Filter(PF)and EKF,the EKF/PF algorithm based on DR/LiDAR integrated combination model is proposed.EKF/PF algorithm by calculating the Mahalanobis distance of two filtering algorithms at each time,the filtering algorithm with smaller Mahalanobis distance is selected as the current state estimation value.The experimental results show that the EKF/PF filter has better robustness than the EKF.Finally,in order to improve the combined positioning accuracy,this paper proposes a DR/LiDAR combined positioning algorithm based on Interactive Multi-Model(IMM)description.First build two motion models of mobile robot moving and stationary,on this basis,through the use of IMM,by calculating the confidence of the two states to weight the two state models,in order to obtain better positioning accuracy.The experimental results show that the IMM-EKF/PF algorithm has a smaller positioning error than the EKF algorithm when the robot has two motion states,moving and stationary.Through the research of this paper,a combined positioning system based on LiDAR and DR is realized,which overcomes the disadvantage that the error accumulates with time when only using DR.Secondly,the EKF/PF algorithm is used in the data fusion stage.Compared with the single filtering algorithm,the positioning performance is improved,and it overcomes the characteristics that the positioning effect is greatly affected by parameter selection when only EKF is used.Finally,using the IMM-EKF/PF algorithm,two kinds of robot motion models are established,which can estimate the position of the robot more accurately when there are two kinds of motion states,moving and stationary.