Structural Lines Aided Monocular Visual-Inertial-Wheel Odometry With Online IMU-Wheel Extrinsic Optimization on ${\mathbb{S}}^{2}$ Manifold

In the article, we focus on the robustness of monocular visual-inertial-wheel odometry (VIWO) in urban environments. In the urban environment, the rapid changes in terrain and lighting intensity are the critical factors that impact the robustness of VIWO. To address the issues mentioned above, this article proposes a novel approach of utilizing structural lines to assist in monocular visual-inertial-wheel odometry, coupled with online optimization of IMU-wheel extrinsic optimization on
$\mathbb{S}^{2}$
manifold. To compensate for the failure of simple point features in strong exposure scenarios, our system incorporates structural line measurements into a sliding-window pose estimator. Moreover, different movements have different observability effects on the extrinsic parameters of sensors. We consider the observability of the extrinsic parameters between IMU-odometers and introduce online optimization of these parameters to improve the robustness of the system. Compared with other VIO and VIWO methods based on point features, experimental results on KAIST's Complex Urban Dataset and campus dataset show that our method has better performance on accuracy and robustness.