ELiOT : End-to-end Lidar Odometry using Transformer Framework

TL;DR

ELiOT introduces an end-to-end transformer-based LiDAR odometry framework that implicitly models motion from sequential point clouds, achieving promising accuracy without traditional geometric methods.

Contribution

It presents a novel transformer architecture for LiDAR odometry that bypasses conventional geometric projections, improving end-to-end learning of motion from point cloud sequences.

Findings

Achieves 7.59% translational error on KITTI dataset.

Achieves 2.67% rotational error on KITTI dataset.

Demonstrates effectiveness of transformer-based approach for LiDAR odometry.

Abstract

In recent years, deep-learning-based point cloud registration methods have shown significant promise. Furthermore, learning-based 3D detectors have demonstrated their effectiveness in encoding semantic information from LiDAR data. In this paper, we introduce ELiOT, an end-to-end LiDAR odometry framework built on a transformer architecture. Our proposed Self-attention flow embedding network implicitly represents the motion of sequential LiDAR scenes, bypassing the need for 3D-2D projections traditionally used in such tasks. The network pipeline, composed of a 3D transformer encoder-decoder, has shown effectiveness in predicting poses on urban datasets. In terms of translational and rotational errors, our proposed method yields encouraging results, with 7.59% and 2.67% respectively on the KITTI odometry dataset. This is achieved with an end-to-end approach that foregoes the need for conventional geometric concepts.