Research On Integrated Navigation Algorithm Based On GNSS/IMU/Vision Multi-Sensor Fusion

Autonomous driving technology has provided convenience for people's lives,such as reducing the incidence of traffic accidents,improving travel efficiency,and reducing labor costs in park operations.Positioning technology is a very important part of autonomous driving technology closely related with the correct judgment of vehicle position and attitude,which is an important basis of path planning,vehicle control and other processes.Therefore,it is essential to ensure the accuracy and stability of positioning.At present,the positioning of autonomous driving vehicles relies mainly on the integration of GNSS(Global Navigation Satellite System)and INS(Inertial Navigation System).Reliable positioning of autonomous driving vehicles can be achieved through the combination of them in most cases.However,there are still two problems not properly solved: 1)GNSS signals introduce delay due to data processing and data transmission;2)GNSS cannot output information when GNSS signals are in occlusion regions,and inertial navigation alone may cause large errors.In this paper,GNSS/INS integrated navigation system was constructed.The delay problem was solved by using the delay estimation and compensation algorithm.At the same time,a visual-inertial odometry system was constructed.Specific research is as follows:Firstly,according to the basic principle of inertial navigation,this paper completed the derivation of inertial navigation algorithm,which was the basis for constructing GNSS/INS integrated navigation system.An IMU(Inertial Measurement Unit)pre-integration algorithm for the fusion of inertial and vision sensors was established,which was the basis for constructing a visual-inertial odometry system.Then the inertial navigation error model was analyzed and the state model of the GNSS/INS integrated navigation system was constructed.The measurement model was also constructed based on the relationship between the measurements and the states.The inertial navigation errors were estimated by Kalman filter algorithm,and the GNSS/INS integrated navigation algorithm was realized and verified by an experiment.Next,in order to remove the delay introduced by GNSS,this paper proposed a delay estimation algorithm based on kinematics,and the delay compensation was carried out by means of residual reconstruction.The algorithms had been verified by a real dataset.Finally,a tightly coupled system of vision and inertial sensors was designed to handle the situation where the GNSS signal was occluded.The visual information was used to limit the accumulation of inertial navigation errors and drift.An experiment was designed to verify it.According to the error analysis in the experiment,GNSS/INS integrated navigation algorithm,delay estimation and compensation algorithm,visual-inertial odometry were stable and effective,which proved to be able to solve the forementioned positioning problems,demonstrating engineering effectiveness of the constructed positioning system.