Influence of IMU’s quality on VIO: based on MSCKF method

Vision inertial odometer (VIO) has been applied for SLAM (Simultaneous Localization and Mapping) this years, such device consists of camera and inertial measurement unit(IMU), and takes advantages of both sensors. When one of sensor’s quality is not as good as the other, we have to trust better one and implement this strategy in our algorithms. In this paper, influences of IMU’s quality on VIO are analyzed, based on ideal IMU’s data and multi-state constraint Kalman filter(MSCKF) method, different levels IMU are implemented in simulation, results show that: for high precision IMU, pure inertial navigation performance is better than VIO; when it comes to medium precision IMU, VIO performance are better if more features are tracked by camera. Further study shows that, the precision of MSCKF method can be improved by adjusting window size.