A Lightweight Machine Learning Pipeline for LiDAR-simulation

TL;DR

This paper introduces a lightweight, learning-based LiDAR simulation method that uses image-to-image translation to produce realistic sensor data, reducing complexity and domain gap in autonomous driving testing.

Contribution

It presents a novel approach that learns sensor behavior from real data and applies image translation to improve LiDAR simulation realism.

Findings

The method generalizes from real to synthetic images effectively.

It reduces the need for complex physics-based LiDAR simulation.

The approach improves the realism of simulated LiDAR data.

Abstract

Virtual testing is a crucial task to ensure safety in autonomous driving, and sensor simulation is an important task in this domain. Most current LiDAR simulations are very simplistic and are mainly used to perform initial tests, while the majority of insights are gathered on the road. In this paper, we propose a lightweight approach for more realistic LiDAR simulation that learns a real sensor's behavior from test drive data and transforms this to the virtual domain. The central idea is to cast the simulation into an image-to-image translation problem. We train our pix2pix based architecture on two real world data sets, namely the popular KITTI data set and the Audi Autonomous Driving Dataset which provide both, RGB and LiDAR images. We apply this network on synthetic renderings and show that it generalizes sufficiently from real images to simulated images. This strategy enables to skip the sensor-specific, expensive and complex LiDAR physics simulation in our synthetic world and avoids oversimplification and a large domain-gap through the clean synthetic environment.