Abstract:LiDAR-based localization and mapping is one of the core components in many modern robotic systems due to the direct integration of range and geometry, allowing for precise motion estimation and generation of high quality maps in real-time. Yet, as a consequence of insufficient environmental constraints present in the scene, this dependence on geometry can result in localization failure, happening in self-symmetric surroundings such as tunnels. This work addresses precisely this issue by proposing a neural network-based estimation approach for detecting (non-)localizability during robot operation. Special attention is given to the localizability of scan-to-scan registration, as it is a crucial component in many LiDAR odometry estimation pipelines. In contrast to previous, mostly traditional detection approaches, the proposed method enables early detection of failure by estimating the localizability on raw sensor measurements without evaluating the underlying registration optimization. Moreover, previous approaches remain limited in their ability to generalize across environments and sensor types, as heuristic-tuning of degeneracy detection thresholds is required. The proposed approach avoids this problem by learning from a collection of different environments, allowing the network to function over various scenarios. Furthermore, the network is trained exclusively on simulated data, avoiding arduous data collection in challenging and degenerate, often hard-to-access, environments. The presented method is tested during field experiments conducted across challenging environments and on two different sensor types without any modifications. The observed detection performance is on par with state-of-the-art methods after environment-specific threshold tuning.

NIDALoc: Neurobiologically Inspired Deep LiDAR Localization

3D LiDAR-Based Global Localization Using Siamese Neural Network

LocNet: Global Localization in 3D Point Clouds for Mobile Vehicles

PointLoc: Deep Pose Regressor for LiDAR Point Cloud Localization

LiDAR-based localization using universal encoding and memory-aware regression

STCLoc: Deep LiDAR Localization with Spatio-Temporal Constraints.

LoD-Loc: Aerial Visual Localization using LoD 3D Map with Neural Wireframe Alignment

Deep Camera Pose Regression Using Pseudo-LiDAR

SGLoc: Scene Geometry Encoding for Outdoor LiDAR Localization

HypLiLoc: Towards Effective LiDAR Pose Regression with Hyperbolic Fusion

Deep Regression for LiDAR-based Localization in Dense Urban Areas

NeRF-Loc: Visual Localization with Conditional Neural Radiance Field.

I2D-Loc: Camera Localization Via Image to LiDAR Depth Flow

DiffLoc: Diffusion Model for Outdoor LiDAR Localization

A novel 3D LiDAR deep learning approach for uncrewed vehicle odometry

Dynamic Scene's Laser Localization by NeuroIV-Based Moving Objects Detection and LiDAR Points Evaluation

LATITUDE: Robotic Global Localization with Truncated Dynamic Low-pass Filter in City-scale NeRF

NeRF-LOAM: Neural Implicit Representation for Large-Scale Incremental LiDAR Odometry and Mapping

Learning-based Localizability Estimation for Robust LiDAR Localization

Fast and Accurate Deep Loop Closing and Relocalization for Reliable LiDAR SLAM

LiDAR Odometry by Deep Learning-Based Feature Points with Two-Step Pose Estimation