OKVIS2: Realtime Scalable Visual-Inertial SLAM with Loop Closure
About
Robust and accurate state estimation remains a challenge in robotics, Augmented, and Virtual Reality (AR/VR), even as Visual-Inertial Simultaneous Localisation and Mapping (VI-SLAM) getting commoditised. Here, a full VI-SLAM system is introduced that particularly addresses challenges around long as well as repeated loop-closures. A series of experiments reveals that it achieves and in part outperforms what state-of-the-art open-source systems achieve. At the core of the algorithm sits the creation of pose-graph edges through marginalisation of common observations, which can fluidly be turned back into landmarks and observations upon loop-closure. The scheme contains a realtime estimator optimising a bounded-size factor graph consisting of observations, IMU pre-integral error terms, and pose-graph edges -- and it allows for optimisation of larger loops re-using the same factor-graph asynchronously when needed.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Visual SLAM | EuRoC (All sequences) | ATE11 | 56 | |
| Visual Odometry | TUM-VI (room sequences) | ATE (cm)5.8 | 31 | |
| Odometry | EuRoC | RMSE ATE (m)0.063 | 12 | |
| Odometry | INOUT-TRANS sequence | ATE Rotation Error (deg)6.976 | 9 | |
| Visual Odometry | TUM-VI Room sequences R1-R6 | Avg RMSE ATE (m)0.067 | 8 | |
| Odometry | TIR-NOISE-IN sequence | ATE Rotation Error (deg)2.712 | 5 | |
| Visual Odometry | Monado SLAM Dataset (MSD) | ATE Success Rate31.2 | 4 |