Abstract
Mobile robots have been widely used in geological exploration, accident rescue, and other fields. As the core technology of these robots, simultaneous localization and mapping is responsible for estimating the robot’s position and building a scene map. With the rapid development of simultaneous localization and mapping technology in multi-sensor fusion, many visual-inertial odometry systems based on camera and inertial measurement unit fusion have been proposed. Although visual-inertial odometry has advantages over visual simultaneous localization and mapping in accuracy and robustness, they still cannot meet the requirements of robots to work stably for a long time. The time misalignment (time offset) between visual and inertial on the timeline is one of the main reasons for that problem. Therefore, this article proposes an online temporal calibration method based on nonlinear optimization, suitable for low-cost, self-assembled stereo visual-inertial odometry systems. We take full advantage of stereo camera, which constructs a new error factor concerning time offset using the epipolar constraint in the stereo camera. The proposed method could reduce the negative influence of features’ velocity errors during temporal calibration and improve the system’s robustness and accuracy. Experiments on the EuRoC data set show that our method’s temporal calibration results are closer to the actual value than the state-of-the-art calibration methods. Our method can significantly improve the system’s accuracy and robustness.
Get full access to this article
View all access options for this article.