Direct LiDAR-Inertial Odometry: Lightweight LIO with Continuous-Time Motion Correction (2203.03749v4)

Abstract: Aggressive motions from agile flights or traversing irregular terrain induce motion distortion in LiDAR scans that can degrade state estimation and mapping. Some methods exist to mitigate this effect, but they are still too simplistic or computationally costly for resource-constrained mobile robots. To this end, this paper presents Direct LiDAR-Inertial Odometry (DLIO), a lightweight LiDAR-inertial odometry algorithm with a new coarse-to-fine approach in constructing continuous-time trajectories for precise motion correction. The key to our method lies in the construction of a set of analytical equations which are parameterized solely by time, enabling fast and parallelizable point-wise deskewing. This method is feasible only because of the strong convergence properties in our nonlinear geometric observer, which provides provably correct state estimates for initializing the sensitive IMU integration step. Moreover, by simultaneously performing motion correction and prior generation, and by directly registering each scan to the map and bypassing scan-to-scan, DLIO's condensed architecture is nearly 20% more computationally efficient than the current state-of-the-art with a 12% increase in accuracy. We demonstrate DLIO's superior localization accuracy, map quality, and lower computational overhead as compared to four state-of-the-art algorithms through extensive tests using multiple public benchmark and self-collected datasets.

Direct LiDAR-Inertial Odometry: Lightweight LIO with Continuous-Time Motion Correction

The paper presents a novel approach to enhancing state estimation and mapping capabilities in mobile robots by addressing motion distortions observed in LiDAR scans during dynamic movements. The authors introduce Direct LiDAR-Inertial Odometry (DLIO), which leverages lightweight algorithms with a focus on efficient and precise motion correction using continuous-time trajectory models. This work illustrates a computationally efficient methodology by incorporating a coarse-to-fine strategy for handling motion distortions, specifically tailored for resource-constrained mobile platforms.

The primary contribution of DLIO is its method of trajectory construction for motion correction. Instead of relying on simplistic models that assume constant velocity, DLIO utilizes a more sophisticated constant jerk and angular acceleration paradigm, allowing for the derivation of analytical equations parameterized by time. This results in parallelizable point-wise deskewing of the LiDAR scans. The nonlinear geometric observer integrated within DLIO aids in initializing the sensitive IMU integration step correctly, enabling robust and accurate state estimation even with computational limitations.

The authors' approach significantly outperforms existing methods both in terms of accuracy and computational efficiency. In particular, DLIO is demonstrated to improve computational efficiency by nearly 20% and localization accuracy by 12% over existing state-of-the-art algorithms. The paper substantiates these claims through evaluations against several benchmarks, such as the Newer College and self-collected datasets. Notably, DLIO excels in environments where aggressive maneuvers or uneven terrain cause significant scan distortion, which is a challenge inadequately addressed by traditional LiDAR odometry and LiDAR-inertial odometry techniques.

From a theoretical standpoint, DLIO introduces a continuous-time method that enhances the precision of motion correction by fitting a smooth trajectory based on time-parameterized equations rather than discrete approximations. This addresses the loss of precision commonly encountered with traditional discrete-time models. Practically, DLIO holds significant implications for improving mobile robot navigation and mapping in unstructured environments, providing more detailed and reliable maps necessary for path planning and other navigational tasks.

Future research developments of DLIO could revolve around its integration and testing in closed-loop flight scenarios, and potentially extending its capabilities with additional features like loop-closure detection to further enhance the robustness of simultaneous localization and mapping (SLAM) systems. The promising results observed motivate further exploration into hierarchical nonlinear observers that can offer similar efficiency benefits across different sensing modalities and environments, thereby broadening the scope of mobile robotics applications in diverse fields.