Multi-source Real-time Precise Positioning Technology For Wheeled Robot

With the rapid development of artificial intelligence and automatic driving industry,wheeled-robots are widely used in consuming services,express logistics,public security patrol,special operations and other fields,and the demand for continuous,real-time,precise and low-cost positioning technology is also more and more urgent.In urban environment,GNSS/INS integrated navigation can provide reliable,continuous and complete high-precision navigation parameters,which is one of the most widely used positioning technologies.However,the low-cost MEMS IMU has huge noise with low accuracy.During the GNSS outages,the positioning error will accumulate rapidly,leading to poorer usability.The GNSS/INS integrated navigation system can be aided by vehicle information of wheeled-robot such as odometer and non-holonomic constraints,which can significantly reduce the drift speed of INS error.However,due to prolonged GNSS outages,the GNSS/INS/Robot-aided integrated positioning system still cannot guarantee the accuracy.Based on the absolute vision positioning of low-cost camera,sparse visual landmarks can provide precise and absolute aiding sources,which establishes a GNSS/INS/Robot-aided/Vision-Landmark multi-source fusion positioning system,that can achieve continuous high-accuracy positioning in GNSS denied environment.With the urgent need of continuous,real-time,precise and low-cost real-time positioning technology for wheeled-robot,this thesis studies the multi-source fusion positioning technology based on GNSS,INS,odometer,vision and other navigation sources,establishes the fusion positioning model,develops the multi-source real-time precise wheeled robot system,and finally verifies the performance by experiment.The main research contents and contributions of this thesis are as follows:1.In order to enhance the robustness of GNSS/INS integrated positioning system and improve the performance of high-precision positioning in GNSS denied environment,a GNSS/INS real-time integrated positioning scheme aided by vehicle information of wheeled-robot is proposed and developed,which called robot real-time integrated positioning scheme.Firstly,based on the classical state and observation model of GNSS/INS integrated navigation,the observation models of dual-antennas heading,wheeled odometer,NHC and ZUPT are established.Then,for real-time requirement,high-accuracy time synchronization technology between GNSS and IMU is developed,and GNSS delay update algorithm is adopted.Finally,the developed integrated positioning module is evaluated and analyzed by real experiment.The experimental results show that the GNSS/INS/Robot-aided integrated positioning system can achieve the initial heading alignment better than 0.5°;in open-sky environment,positioning accuracy is centimeter level,and heading accuracy is better than 0.12°;during GNSS outages of 30 seconds,the integrated system can maintain the horizontal positioning accuracy better than 0.3m.2.In order to further improve the continuity of high-accuracy positioning of GNSS/INS/Robot-aided integrated positioning system in urban environment,a robot integrated positioning scheme based on visual landmark is proposed and developed.Firstly,the space-time synchronization framework of camera and IMU is established.Then,the observation models of Visual/INS loose coupled and tight coupled are established respectively.Then,the artificial visual landmark is designed,with the algorithm of feature extraction and matching.Finally,with real data sampled by wheeled robot,the quantitative evaluation and analysis of the observation effect of the visual fusion algorithm are carried out.The experimental results show that the performance during GNSS outages can be significantly improved whether the loose-coupled or tight-coupled algorithm,but the Visual/INS tight-coupled algorithm has higher accuracy and better effect.During GNSS outages of 20 seconds with continuous visual-landmark aiding of about 5 seconds,using Visual/INS tight coupled algorithm,the GNSS/INS/Robot-aided/Visual-landmark multi-source fusion positioning system can maintain a horizontal positioning accuracy better than 0.18 m.3.Based on the theory and method of the robot integrated positioning above,a multi-source real-time precise positioning wheeled-robot system is developed.Firstly,based on GNSS/INS real-time integrated positioning technology aided by the vehicle information,INS-Probe,a real-time integrated positioning module,is developed.Then,taking INS-Probe as the core,the GNSS/INS/Robot-aided/Visual-landmark multi-source fusion positioning system for wheeled-robot is established by using the Visual/INS tight-coupled algorithm based on visual landmark.Finally,the real-time performance of the integrated navigation module and the multi-source fusion positioning system are evaluated in urban environment.The experimental results show that the multi-source fusion positioning system has integral structure,and the real-time multi-source fusion positioning software runs steadily,which can improve the continuous real-time precise positioning for wheeled robot.In conclusion,for the positioning and navigation requirements of wheeled robot,this thesis makes an in-depth research on GNSS/INS/Robot-aided/Visual-landmark multi-source fusion positioning technology,establishes a complete filtering model,develops a GNSS/INS/Robot-aided real-time integrated positioning module,and develops a multi-source fusion positioning robot platform based on INS-Probe,and comprehensively evaluates and analyzes the algorithm model and hardware system by experiment.The research in this thesis provides a reliable references scheme for the continuous,real-time and precise positioning of wheeled robot in urban environment.