Fast LiDAR Data Generation with Rectified Flows

TL;DR

This paper introduces R2Flow, a fast and efficient LiDAR data generator using rectified flows and Transformer architecture, reducing sampling steps while maintaining high fidelity for autonomous robotics applications.

Contribution

The paper proposes R2Flow, a novel LiDAR generative model that significantly speeds up data generation with fewer sampling steps using rectified flows and a Transformer-based architecture.

Findings

R2Flow achieves high-quality LiDAR data generation with fewer sampling steps.

The model demonstrates efficiency and effectiveness on the KITTI-360 dataset.

Results show improved generation speed without sacrificing quality.

Abstract

Building LiDAR generative models holds promise as powerful data priors for restoration, scene manipulation, and scalable simulation in autonomous mobile robots. In recent years, approaches using diffusion models have emerged, significantly improving training stability and generation quality. Despite their success, diffusion models require numerous iterations of running neural networks to generate high-quality samples, making the increasing computational cost a potential barrier for robotics applications. To address this challenge, this paper presents R2Flow, a fast and high-fidelity generative model for LiDAR data. Our method is based on rectified flows that learn straight trajectories, simulating data generation with significantly fewer sampling steps compared to diffusion models. We also propose an efficient Transformer-based model architecture for processing the image representation of LiDAR range and reflectance measurements. Our experiments on unconditional LiDAR data generation using the KITTI-360 dataset demonstrate the effectiveness of our approach in terms of both efficiency and quality.