R-LIO: Rotating Lidar Inertial Odometry and Mapping

Abstract

In this paper, we propose a novel simultaneous localization and mapping algorithm, R-LIO, which combines rotating multi-line lidar and inertial measurement unit. R-LIO can achieve real-time and high-precision pose estimation and map-building. R-LIO is mainly composed of four sequential modules, namely nonlinear motion distortion compensation module, frame-to-frame point cloud matching module based on normal distribution transformation by self-adaptive grid, frame-to-submap point cloud matching module based on line and surface feature, and loop closure detection module based on submap-to-submap point cloud matching. R-LIO is tested on public datasets and private datasets, and it is compared quantitatively and qualitatively to the four well-known methods. The test results show that R-LIO has a comparable localization accuracy to well-known algorithms as LIO-SAM, FAST-LIO2, and Faster-LIO in non-rotating lidar data. The standard algorithms cannot function normally with rotating lidar data. Compared with non-rotating lidar data, R-LIO can improve localization and mapping accuracy in rotating lidar data.

LiDAR-based SLAM for robotic mapping: state of the art and new frontiers

“…This framework used segmented point cloud data from a spinning-actuated LiDAR in a concurrent multithreaded matching pipeline to estimate 6D pose with a high update rate and low latency. Chen et al (2022) developed R-LIO (rotating LiDAR inertial odometry), a novel SLAM algorithm that integrated a spinning-actuated 3D LiDAR with an IMU. R-LIO was capable of high-precision, real-time position estimation and map construction.…”

Anchor-Based, Real-Time Motion Compensation for High-Resolution mmWave Radar

“…Meanwhile, LiDAR and visual-inertial odometry systems often rely upon IMU fusion [60], [61], [62] and/or computationally intensive correction algorithms [63], [64], yielding the same downsides as radar solutions. The authors in [64], for instance, rely upon a pre-defined quadratic motion model for platform vibration within each chirp interval, adopting a two-step, semi-iterative method to optimally estimate and eliminate both the primary and quadratic vibration components.…”

Lightweight High-Bandwidth LiDAR-Inertial Odometry Employing Sparse Incremental Hash-Voxels

“…Despite their advantages, hash-voxel face challenges like hash conflicts during cloud insertion. Bai's work [15] addressed this with a dynamic hash-voxel employing conservative insertion and passive deletion, achieving faster point cloud insertion and NN searching speed than ikd-tree.…”

“…The introduction of the incremental kd-tree [14] in these frameworks notably reduces computational time and memory usage compared to traditional kd-tree and other tree-like structures. However, the inherent time complexity of strict nearest-neighbor (NN) search in the ikd-tree may be deemed unnecessary [15]. In pursuit of lightweight solutions, efficient management of sparse point cloud maps is crucial.…”