LINS: A Lidar-Inertial State Estimator for Robust and Efficient Navigation (1907.02233v3)

Abstract: We present LINS, a lightweight lidar-inertial state estimator, for real-time ego-motion estimation. The proposed method enables robust and efficient navigation for ground vehicles in challenging environments, such as feature-less scenes, via fusing a 6-axis IMU and a 3D lidar in a tightly-coupled scheme. An iterated error-state Kalman filter (ESKF) is designed to correct the estimated state recursively by generating new feature correspondences in each iteration, and to keep the system computationally tractable. Moreover, we use a robocentric formulation that represents the state in a moving local frame in order to prevent filter divergence in a long run. To validate robustness and generalizability, extensive experiments are performed in various scenarios. Experimental results indicate that LINS offers comparable performance with the state-of-the-art lidar-inertial odometry in terms of stability and accuracy and has order-of-magnitude improvement in speed.

Overview of LINS: A Lidar-Inertial State Estimator for Robust and Efficient Navigation

The paper introduces LINS, a Lidar-Inertial Navigation System designed to enhance real-time ego-motion estimation for unmanned ground vehicles (UGVs). Leveraging a tightly-coupled sensor fusion approach, the system incorporates a 6-axis Inertial Measurement Unit (IMU) and a 3D lidar to provide robust navigation capabilities even in challenging environments. The authors employ an iterated error-state Kalman filter (ESKF) implemented within a novel robocentric framework, advancing beyond existing methodologies by significantly reducing computational load.

Methodology and Contributions

The primary innovative aspect of LINS is its use of the iterated ESKF within a robocentric formulation, where motion estimation errors are continuously corrected by establishing new feature correspondences in each iteration. This approach mitigates the filter divergence typically encountered in long-duration operations by updating the state in a locally shifting reference frame.

Key contributions of this research include:

1. Development of a lidar-inertial odometry algorithm that surpasses previous implementations by achieving an order-of-magnitude improvement in computational efficiency, thereby rendering the approach feasible for real-time applications.
2. Introduction of a robocentric iterated ESKF, validated in diverse scenarios, to ensure both high performance and stability when compared to existing state-of-the-art solutions.
3. Public accessibility of the source code, offering a groundbreaking tightly-coupled LIO solution leveraging iterated Kalman filtering to solve the 6 DOF ego-motion problem.

Experimental Evaluation

The performance assessment of LINS involved extensive testing across varied scenarios, including urban, port, industrial park, forest, and indoor environments. Several critical insights were reported:

• Accuracy: The LINS system demonstrates comparable precision to state-of-the-art lidar-inertial odometry systems, with measured relative motion drift consistently within competitive boundaries.
• Efficiency: The LIO module reduced processing times to under 30 milliseconds per scan, a stark contrast to the 100+ milliseconds observed with the state-of-the-art systems based on graph optimization.
• Robustness: When evaluated against feature-sparse environments, LINS maintained its stability, outperforming lidar-only methods in east-to-align trajectory tasks.

Results highlight LINS’s ability to generate reliable and drift-resilient trajectory estimates, even in scenarios traditionally challenging for lidar-only systems.

Implications and Future Work

The practical utility of LINS lies in its real-time application potential due to its computational tractability, a critical requirement for mobile robotic applications engaged in dynamic and feature-deficient environments. By providing a robust mechanism for effectively fusing IMU and lidar data, LINS stands to inform the development of advanced navigation systems, augmenting the performance of UGVs in complex, unpredictable settings.

Future directions may explore enriching the feature extraction processes or integrating machine learning techniques to enhance system resilience and adaptability further. Additionally, extending the field trials to include varied domain-specific scenarios could push the boundaries of its applicability, fostering advancements in autonomous vehicle technologies beyond ground-based platforms.

In conclusion, LINS represents a significant step forward in the efficient fusion of lidar and inertial data, embodying a confluence of accuracy, robustness, and computational speed, aligning well with the demands of next-generation autonomous systems.