Measurement-Update-Delayed-MSCKF Driven Reliable Real-Time PPP/INS/Visual Integration

Reliable real-time positioning, a basic requirement for intelligent applications, has been focused on by researchers and engineering applications along with the developments of intelligent applications like intelligent agriculture and intelligent transportations. Currently, the widely used positioning technique is these Global Navigation Satellite System (GNSS)-based methods. However, the continuity and reliability of these GNSS-based positioning methods would be degraded significantly while suffering from frequent satellite signal outages. To overcome such a drawback of the GNSS positioning system, this work tries to provide a real-time reliable positioning algorithm by integrating the GNSS Precise Point Positioning (PPP), Inertial Navigation System (INS), and vision. This method will significantly upgrade the accuracy, continuity, and reliability of GNSS real-time positioning solutions in complex urban environments. In such a method, the Multi-State Constraint Kalman Filter (MSCKF) is adopted to integrate the PPP and the Visual-Inertial Odometry (VIO). However, due to the different intervals and different time delays of GNSS, INS, and visual image in data preprocessing (i.e., decoding), it is difficult to receive these data in the microcomputer processor at the same time, which prevents the operation of the measurement-update of the MSCKF in real-time. To solve such a problem, this work provides a measurement-update-delayed MSCKF, in which the delayed real-time multi-source data can be integrated accurately without significant positioning accuracy degradation. Meanwhile, the IGGIII-robust function is introduced into such a real time MSCKF to mitigate the impact of those abnormal GNSS observations on positioning accuracy. Results based on a 1.67-hours typical urban dataset showed that (1) the positioning accuracies of the proposed algorithm in terms of Root Mean Square Error (RMSE) were 0.31 m, 0.41 m, and 0.81 m in the east, north, and up directions, respectively, and 94% horizontal positioning error was within 0.3 m; (2) the measurement-update-delayed MSCKF effectively eliminates the influences of time delays of GNSS and visual measurement on real time positioning, which brings a higher positioning accuracy; (3) the robust MSCKF significantly reduces the abnormal positioning errors, and limits the maximum positioning errors within 5 meters.