MM-LINS: a Multi-Map LiDAR-Inertial System for Over-Degenerate Environments

TL;DR

The paper tackles SLAM failures caused by over-degeneracy in dynamic, cluttered environments by proposing MM-LINS, a multi-map LiDAR–inertial system. It combines a degeneracy-aware front-end with dynamic initialization and a back-end that uses inter-map similarity via Scan Context to fuse sleeping maps with the current active map, delivering high-precision, globally consistent trajectories. The key contributions are the first multi-map LiDAR–inertial SLAM framework, a robust degeneracy-detection mechanism with dynamic initialization, and a constraint-enhanced map fusion strategy validated on both public degenerate datasets and real-world scenarios. The approach demonstrates improved robustness and accuracy in challenging conditions such as crowds, obstruction by debris, and smoke, with open-source code to enable broader adoption.

Abstract

SLAM plays a crucial role in automation tasks, such as warehouse logistics, healthcare robotics, and restaurant delivery. These scenes come with various challenges, including navigating around crowds of people, dealing with flying plastic bags that can temporarily blind sensors, and addressing reduced LiDAR density caused by cooking smoke. Such scenarios can result in over-degeneracy, causing the map to drift. To address this issue, this paper presents a multi-map LiDAR-inertial system (MM-LINS) for the first time. The front-end employs an iterated error state Kalman filter for state estimation and introduces a reliable evaluation strategy for degeneracy detection. If over-degeneracy is detected, the active map will be stored into sleeping maps. Subsequently, the system continuously attempts to construct new maps using a dynamic initialization method to ensure successful initialization upon leaving the over-degeneracy. Regarding the back-end, the Scan Context descriptor is utilized to detect inter-map similarity. Upon successful recognition of a sleeping map that shares a common region with the active map, the overlapping trajectory region is utilized to constrain the positional transformation near the edge of the prior map. In response to this, a constraint-enhanced map fusion strategy is proposed to achieve high-precision positional and mapping results. Experiments have been conducted separately on both public datasets that exhibited over-degenerate conditions and in real-world environments. These tests demonstrated the effectiveness of MM-LINS in over-degeneracy environment. Our codes are open-sourced on Github.

Figures (8)

• Figure 1: Left Column: Satellite map and trajectory of the robot during the real-world campus dataset. LiDAR obstruction caused by floating garbage bags is illustrated through real images. Demonstration implementation of our algorithm. Right Column: Point cloud maps and trajectories created using SOTA algorithms and ours.
• Figure 2: MM-LINS overview.
• Figure 3: The conceptual overview of the map merging idea. The internal constraints include both loop closure constraints and odometry constraints, which are respectively denoted by dashed and solid lines. Enhanced constraints comprise those detected by similarity measures.
• Figure 4: A visual demonstration of degeneracy event and map merging.
• Figure 5: Performance comparison of different algorithms on the Factory-1 dataset. The subfigures depict the following: (a) Real-world operation of the handheld device and the excellent performance of our algorithm in this case, (b) - (e) Point cloud maps and trajectories of SOTA algorithms in the same situation.