Super Odometry: IMU-centric LiDAR-Visual-Inertial Estimator for Challenging Environments (2104.14938v2)

Abstract: We propose Super Odometry, a high-precision multi-modal sensor fusion framework, providing a simple but effective way to fuse multiple sensors such as LiDAR, camera, and IMU sensors and achieve robust state estimation in perceptually-degraded environments. Different from traditional sensor-fusion methods, Super Odometry employs an IMU-centric data processing pipeline, which combines the advantages of loosely coupled methods with tightly coupled methods and recovers motion in a coarse-to-fine manner. The proposed framework is composed of three parts: IMU odometry, visual-inertial odometry, and laser-inertial odometry. The visual-inertial odometry and laser-inertial odometry provide the pose prior to constrain the IMU bias and receive the motion prediction from IMU odometry. To ensure high performance in real-time, we apply a dynamic octree that only consumes 10 % of the running time compared with a static KD-tree. The proposed system was deployed on drones and ground robots, as part of Team Explorer's effort to the DARPA Subterranean Challenge where the team won $1^{st}$ and $2^{nd}$ place in the Tunnel and Urban Circuits, respectively.

Super Odometry: An Advanced IMU-centric Estimator for Challenging Environments

The academic paper titled "Super Odometry: IMU-centric LiDAR-Visual-Inertial Estimator for Challenging Environments" examines the development of a novel simultaneous localization and mapping (SLAM) system. Designed to tackle the unique challenges presented by subterranean and other perceptually-degraded environments, this paper presents an Inertial Measurement Unit (IMU)-centric sensor fusion pipeline that integrates tightly-coupled and loosely-coupled approaches to achieve robust performance in GPS-denied scenarios.

Core Contributions and Implementation

1. IMU-centric Sensor Fusion Pipeline: The authors introduce an IMU-centric approach that organizes IMU data as the primary sensor source. This is predicated on the IMU's ability to provide smooth and high-fidelity measurements with minimal outliers, although subject to bias drift. By integrating LiDAR and visual sensor inputs to constrain this bias, the pipeline achieves accurate real-time state estimates in environments characterized by obscurants like dust and fog.
2. Fusion Methodology: Super Odometry combines the strengths of loosely and tightly-coupled fusion methods. The tightly-coupled approach ensures accuracy through robust performance with the integration of multi-sensor inputs, while the loosely-coupled approach enhances resilience by permitting multiple estimation engines, reducing the risk of sensor failure.
3. Dynamic Octree Implementation: To address computational inefficiencies, the paper proposes using a dynamic octree method for better handling and organizing 3D point data. This significantly improves real-time performance in scan matching processes by efficiently managing data associations, compared to the traditional KD-tree method.
4. Real-World Deployment and Evaluation: The proposed methodology has been deployed across different robotic platforms including drones and ground robots. Extensive evaluations were conducted in challenging scenarios, accounting for aggressive motions and visually/geometrically challenging settings.

Numerical Results and Observations

The paper documents the comparative performance of Super Odometry with other baseline methods (e.g., LOAM, LIO-SAM, VINS) across various datasets. Remarkably, Super Odometry demonstrates superior accuracy, with notably lower Absolute Trajectory Error (ATE) across test sequences involving low light, air-born particles, long corridors, and subterranean shafts. These results underscore the robustness of the fusion approach, illustrating consistent and resilient state estimation.

Practical and Theoretical Implications

The implications of this research are significant for robotics applications that operate in environments lacking perceptual features or subjected to challenging atmospheric conditions. By ensuring robust state estimation while efficiently integrating multiple sensors, Super Odometry enhances autonomous navigation and situational awareness, which is crucial for applications in search and rescue operations, industrial inspections, and exploration missions in GPS-denied environments.

From a theoretical perspective, this work advances the field of SLAM by illustrating the benefits of an IMU-centric approach. It prompts further exploration into multi-modal sensor fusion strategies that prioritize environmental resilience and accuracy in state estimation.

Future Developments

The integration of additional environmental-independent sensors such as thermal imaging or radar presents promising avenues for future research. Moreover, advancing real-time processing algorithms to further reduce computational overheads and enhance scalability on embedded systems would be worthwhile investigations.

In summary, the paper delivers a comprehensive account of a sophisticated SLAM system tailored for challenging environments, reinforcing the importance of sensor fusion in autonomous robotics and paving the way for further innovation in the domain.