CEAS EuroGNC 2022
|
Visual Odometry Fusion with GNSS/IMU Localization of UAVs in Urban Areas and Integrity Monitoring
|
| Mats Martens |
Research Assistant, Technische Universität Berlin, Chair of Flight Guidance and Air Transport, 10587, Berlin, Germany. |
|
| Abstract:
Urban environments represent a challenging area for UAVs as part of Urban Air Mobility. The reliance on the conventional fusion of GNSS and INS may not fulfill the higher navigation specifications, which could be issued by authorities due to the increased risk of urban air operation. Within this paper, GNSS measurements are fused with those of a consumer grade IMU and VO derived measurements of a monocular camera setup. The intention is to monitor the integrity of the filter solutions and to increase the continuity compared to the conventional fusion. For that purpose, a novel filter architecture is presented, comprising an INS/GNSS and an INS/VO Extended Kalman Filter running in parallel. Using integrity monitoring strategies, unreliable, redundant pseudo-range measurements can be rejected within the INS/GNSS filter through Innovation Filtering. Additionally, the integrity of both filters is monitored using a chi-square test statistic and a H0 hypothesis test. Once the INS/GNSS filter solution is detected as unreliable, the INS/VO filter bridges the outage until the INS/GNSS filter recovered. By fusing the IMU measurements with those of the VO algorithm, the drift can be drastically reduced. Finally, the proposed filter architecture is applied to a recorded test dataset and the assumed benefits are verified. While the accuracy of the parallel design is proven to be qualitatively better than the individual filters alone, the proposed filter doubles the continuity property of the navigation solution.
|
| Keywords: Visual Odometry; Integrity Monitoring; UAV Navigation; Sensor Fusion; Kalman Filter |
 View PDF CEAS-GNC-2022-016 |