Study On Finite Impulse Response Filtering Algorithm For INS/Vision Integrated Navigation

In recent years,with the advancement of technology,an increasing number of intelligent robots have entered people’s lives.They help people complete various tasks and have demonstrated great value in many industries.Mobile robots can perform precise work in complex environments outside the laboratory.Navigation technology that integrates multiple disciplines such as automatic control,computer,microelectronics,optics,mechanics,and mathematics is the foundation.Therefore,robot navigation has gradually become a research hotspot in this field.The most well-known navigation technology among existing navigation technologies is the Global Navigation Satellite System(GNSS).Although it can provide accurate geographic location,vehicle speed,and precise time information,in indoor and underground environments with building obstructions,satellite signals are severely weakened due to the influence of blocking objects,leading to decreased accuracy and stability of GNSS navigation.In order to solve this problem,this article studied the filtering algorithm for indoor robot navigation.We designed an inertial(Inertial Navigation System,INS)/visual combined navigation model with a limited pulse response,built an INS/visual combination navigation experimental platform,proposed an adaptive bidirectional optimal limited impulse response filtering(Unbiased Finite Impulse Response Filtering,UFIR)algorithm and a multi-frequency UFIR filtering algorithm,and designed experiments to verify the robustness of the algorithm.The main contents are as follows:Firstly,four mainstream navigation technologies were analyzed: GNSS uses navigation satellites for positioning,which has the advantages of high accuracy,fast response speed,and all-weather availability.However,its performance is poor in environments with obstructions such as indoors.Visual navigation obtains its own position by identifying feature coordinates in images,but its navigation performance is affected by image quality.INS obtains its pose through its own gyroscope and accelerometer,with strong autonomy,but the solution error of the Inertial Measurement Unit(IMU)will accumulate over time,resulting in low long-term navigation accuracy.Li DAR obtains information about surrounding obstacles by receiving reflected laser beams to achieve navigation positioning,with high positioning accuracy.However,situations such as transparent glass,metal objects,or lack of sufficient features in the navigation environment can reduce the navigation accuracy of Li DAR.In order to overcome the shortcomings of single navigation technology and meet the demand for continuous and stable navigation information,this paper proposes INS/visual combined navigation,fully leveraging the advantages of INS and visual navigation to improve navigation accuracy and robustness.Secondly,an INS/visual combined navigation experimental platform was designed and built.The experimental platform consists of two parts: hardware and software.The hardware part mainly includes IMU,binocular camera,and mobile robot,etc.The software part is a visualization control software developed in C++ language,which can realize the functions of data acquisition,storage,and display for inertial sensors and visual sensors.The experimental platform lays the foundation for the performance verification of the algorithms proposed in this paper.Based on the combination navigation data fusion filter of Finite Impulse Response(FIR)structure,in order to further improve navigation accuracy,this paper proposes an adaptive bidirectional UFIR filtering algorithm based on INS/visual combined navigation system.This algorithm deduces a UFIR algorithm similar to Kalman form on the basis of traditional batch processing UFIR filtering algorithm.Meanwhile,the improved UFIR algorithm is used for forward and backward filtering respectively,and the Mahalanobis distance is introduced as a reference to describe the filtering effect,achieving adaptive adjustment of filtering window length.Experimental results show that the proposed adaptive bidirectional UFIR filtering algorithm can effectively improve the positioning accuracy of mobile robots.Aiming at the problem of inconsistent sensor sampling frequencies in INS/visual combined navigation system,this paper proposes a one-step-ahead prediction enhanced multifrequency UFIR filtering algorithm.Due to the physical characteristics of sensors,the sampling frequency of inertial sensors is often faster than that of vision sensors in practical process.The one-step-ahead prediction enhanced multi-frequency UFIR filtering algorithm separates the prediction update from the measurement update,using the sampling frequency of inertial sensors as the basis.When there is no visual information,only the prediction update is carried out,and when the visual information appears,the measurement update is performed.Experimental results show that the traditional data fusion algorithm for combined navigation has a significant impact on data fusion accuracy due to data missing caused by slow data fusion cycle,which can be effectively reduced in the proposed one-step-ahead prediction enhanced multi-frequency UFIR filtering algorithm.In summary,this paper collects the pose data of mobile robots using the INS/visual combined navigation experimental platform and designs two data fusion filtering algorithms for different requirements.The proposed adaptive bidirectional UFIR filtering algorithm improves the robustness and accuracy of robot navigation.The one-step-ahead prediction enhanced multi-frequency UFIR filtering algorithm solves the problem of multi-frequency between different sensors in the process of combined navigation.Compared with traditional KF and FIR,this algorithm has higher robustness and accuracy.Overall,this study demonstrates the effectiveness of the proposed algorithms in improving the performance of INS/visual combined navigation system.