VICO: Visual-Inertial Continuous-Time Odometry Based on Generalized Hermite Spline

Continuous-time simultaneous localization and mapping (SLAM) facilitates the seamless fusion of asynchronous and high update-rate sensors. The traditional continuous-time parameterizaton adopts the cumulative $B$-splines, causing the complexity in implementation and abstraction of control points. To address these issues, this paper proposes a continuous-time visual-inertial odometry (VIO) based on generalized Hermite spline. The analytical temporal derivatives and Jacobians with respect to the control points are derived, so that the VIO is further formulated as a sliding window based optimization. To validate the efficacy of the proposed method, the extensive evaluations are conducted on the TUM VI and EuRoC dataset. The results demonstrate the state-of-the-art accuracy and realtime performance. Our implementation is fully open-source at https://github.com/FALCONS-Lab/VICO.