Shafiezadeh, Ali (University of Alberta), Bhatt, Neel P. (The University of Texas at Austin), Hashemi, Ehsan (University of Alberta)

LiDAR-Based Navigation Using Normal Distributions Transform Filter

Scheduled for presentation during the Regular Session "LiDAR-based perception" (FrBT6), Friday, September 27, 2024, 14:50−15:10, Salon 14

2024 IEEE 27th International Conference on Intelligent Transportation Systems (ITSC), September 24- 27, 2024, Edmonton, Canada

This information is tentative and subject to change. Compiled on December 26, 2024

Abstract

Model uncertainties and dynamic objects are the main challenges for LiDAR-based state estimation in perception systems for autonomous mobile robots/vehicles navigation in unstructured environments. A novel state observer, which integrates point cloud matching updates with a motion model through an optimal variance filter, is designed and experimentally verified in various scenes in urban settings and unstructured environments. The designed state observer addresses the inaccuracies found in environments with poor visibility, such as those with heavy vegetation and dynamic objects, and feature-less scenarios which are challenging for existing LiDAR-based pose estimation methods in autonomous navigation. The detectability of the proposed dynamical process which guarantees the stability of the estimation error dynamics is also studied. We have evaluated the state estimator’s performance, in terms of estimation accuracy and processing time in complex urban navigation tasks, and compared it with existing approaches and baselines.