COIN-LIO: Complementary Intensity-Augmented LiDAR Inertial Odometry (2310.01235v4)

Abstract: We present COIN-LIO, a LiDAR Inertial Odometry pipeline that tightly couples information from LiDAR intensity with geometry-based point cloud registration. The focus of our work is to improve the robustness of LiDAR-inertial odometry in geometrically degenerate scenarios, like tunnels or flat fields. We project LiDAR intensity returns into an intensity image, and propose an image processing pipeline that produces filtered images with improved brightness consistency within the image as well as across different scenes. To effectively leverage intensity as an additional modality, we present a novel feature selection scheme that detects uninformative directions in the point cloud registration and explicitly selects patches with complementary image information. Photometric error minimization in the image patches is then fused with inertial measurements and point-to-plane registration in an iterated Extended Kalman Filter. The proposed approach improves accuracy and robustness on a public dataset. We additionally publish a new dataset, that captures five real-world environments in challenging, geometrically degenerate scenes. By using the additional photometric information, our approach shows drastically improved robustness against geometric degeneracy in environments where all compared baseline approaches fail.

• The paper introduces an intensity-enhanced odometry pipeline that leverages LiDAR intensity images to bolster registration in degenerate settings.
• It proposes a novel feature selection scheme that integrates photometric error minimization with an iterated EKF to improve robustness.
• The authors provide the ENWIDE dataset, offering real-world sequences from challenging environments to validate enhanced odometry performance.

An Expert Overview of COIN-LIO: Complementary Intensity-Augmented LiDAR Inertial Odometry

The paper, "COIN-LIO: Complementary Intensity-Augmented LiDAR Inertial Odometry," introduces a LiDAR Inertial Odometry (LIO) pipeline that integrates LiDAR intensity information with geometry-based point cloud registration. The focus of this research is improving the robustness of LIO in geometrically degenerate environments such as tunnels, flat fields, or planar environments, where existing LiDAR-inertial odometry techniques face significant challenges.

Key Contributions

1. Intensity-Enhanced Odometry Pipeline: The authors propose a novel method to utilize LiDAR intensity returns by transforming them into intensity images, which undergo a specialized image processing pipeline to enhance brightness consistency. This approach effectively suppresses noise and improves the utility of intensity as an additional data modality in scenarios where geometric data alone is insufficient.
2. Novel Feature Selection Scheme: A significant innovation of this paper is the feature selection strategy that identifies uninformative directions in point cloud registration and selects image patches offering complementary information. This method utilizes photometric error minimization fused with inertial measurements, operating within an iterated Extended Kalman Filter (iEKF) framework.
3. Introduction of the ENWIDE Dataset: The paper acknowledges a dearth of datasets focusing on geometrically degenerate environments and addresses this by providing the ENWIDE dataset. This dataset encompasses real-world scenarios, capturing sequences in environments like tunnels and open fields, thus providing a valuable resource for further research.

Results and Implications

Numerical results from the standard dataset indicate enhanced accuracy over competing methodologies, highlighting the pipeline's resilience in environments typically provoking existing systems to falter. The ENWIDE dataset further establishes COIN-LIO's robustness, exhibiting substantial performance improvements where baseline systems fail.

This paper's findings suggest practical implications in the development of autonomous systems capable of operating in challenging environments, such as underground mining operations, urban canyons, or densely forested areas. The utilization of LiDAR intensity information emerges as a viable strategy to augment existing odometry frameworks, particularly in situations where GNSS is unreliable or unavailable.

Future Directions

The research opens avenues for further exploration into the fusion of multi-modal sensory data to enhance robotic perception systems. Future investigations could refine the photometric processing pipeline to handle diverse LiDAR configurations better or explore machine learning techniques to dynamically adapt feature selection strategies. Additionally, the dissemination of the ENWIDE dataset is poised to catalyze advancements in the robustness of VIO and LIO systems, encouraging further exploration into this challenging aspect of robotics.

In conclusion, COIN-LIO presents a compelling advancement in LiDAR inertial odometry, with significant implications for the robustness of autonomous navigation in geometrically challenging environments. The release of the ENWIDE dataset will likely support ongoing research efforts and inspire continued innovation in this critical area of mobile robotics.