KISS-SLAM: A Simple, Robust, and Accurate 3D LiDAR SLAM System With Enhanced Generalization Capabilities

TL;DR

This work tackles robust, real-time LiDAR-only SLAM with minimal parameter tuning across diverse sensors and motion profiles. It introduces KISS-SLAM, a pipeline combining KISS-ICP odometry, local map construction, loop closure using 2D density images and ORB features, and pose-graph optimization, yielding accurate and globally consistent trajectories. The authors demonstrate state-of-the-art pose accuracy across multiple datasets, generalization to different sensors, and practical usefulness of the 3D maps for navigation, all while maintaining a simple, modular design and real-time performance. The open-source release provides a strong baseline for LiDAR SLAM and a flexible starting point for future extensions.

Abstract

Robust and accurate localization and mapping of an environment using laser scanners, so-called LiDAR SLAM, is essential to many robotic applications. Early 3D LiDAR SLAM methods often exploited additional information from IMU or GNSS sensors to enhance localization accuracy and mitigate drift. Later, advanced systems further improved the estimation at the cost of a higher runtime and complexity. This paper explores the limits of what can be achieved with a LiDAR-only SLAM approach while following the "Keep It Small and Simple" (KISS) principle. By leveraging this minimalistic design principle, our system, KISS-SLAM, archives state-of-the-art performances in pose accuracy while requiring little to no parameter tuning for deployment across diverse environments, sensors, and motion profiles. We follow best practices in graph-based SLAM and build upon LiDAR odometry to compute the relative motion between scans and construct local maps of the environment. To correct drift, we match local maps and optimize the trajectory in a pose graph optimization step. The experimental results demonstrate that this design achieves competitive performance while reducing complexity and reliance on additional sensor modalities. By prioritizing simplicity, this work provides a new strong baseline for LiDAR-only SLAM and a high-performing starting point for future research. Further, our pipeline builds consistent maps that can be used directly for further downstream tasks like navigation. Our open-source system operates faster than the sensor frame rate in all presented datasets and is designed for real-world scenarios.

Figures (3)

• Figure 1: Our proposed SLAM system can accurately estimate the trajectory for a drive of more than 9 km length in real-time. Using the same parameter configuration, we achieve a similar output for two different LiDAR sensors with different scan patterns and resolutions from the HeLiPR dataset jung2024ijrr. In both cases, our system can compute the odometry, successfully find loop closures across large time spans, and output a globally consistent trajectory.
• Figure 2: Overview of KISS-SLAM. Our pipeline processes each LiDAR scan and first computes the odometry of the scanner, see Sec. \ref{['sec:odom']}. Successively, the system integrates the LiDAR point cloud with the motion estimate into a local map in Sec. \ref{['sec:mapping']}. When the LiDAR moton exceeds $\beta$ meters, the system updates the pose graph and searches for loop closures, see Sec. \ref{['sec:closure']}. If they are positively validated based on overlap criteria, we perform a pose graph optimization step.
• Figure 3: Occupancy grid maps generated with our approach. The top shows the 3D occupancy grid we generate using KISS-SLAM, after processing all the 3D scans from an Hesai XT-32 with a Clearpath Husky. The bottom displays the corresponding 2D grid generated by slicing the 3D map. We then use this map to do 2D localization using the approach of Grisetti et al. grisetti2018github.