Bridging the Domain Gap between Synthetic and Real-World Data for Autonomous Driving

Xiangyu Bai,Yedi Luo,Le Jiang,Aniket Gupta,Pushyami Kaveti,Hanumant Singh,Sarah Ostadabbas

2023-06-05

Abstract:Modern autonomous systems require extensive testing to ensure reliability and build trust in ground vehicles. However, testing these systems in the real-world is challenging due to the lack of large and diverse datasets, especially in edge cases. Therefore, simulations are necessary for their development and evaluation. However, existing open-source simulators often exhibit a significant gap between synthetic and real-world domains, leading to deteriorated mobility performance and reduced platform reliability when using simulation data. To address this issue, our Scoping Autonomous Vehicle Simulation (SAVeS) platform benchmarks the performance of simulated environments for autonomous ground vehicle testing between synthetic and real-world domains. Our platform aims to quantify the domain gap and enable researchers to develop and test autonomous systems in a controlled environment. Additionally, we propose using domain adaptation technologies to address the domain gap between synthetic and real-world data with our SAVeS$^+$ extension. Our results demonstrate that SAVeS$^+$ is effective in helping to close the gap between synthetic and real-world domains and yields comparable performance for models trained with processed synthetic datasets to those trained on real-world datasets of same scale. This paper highlights our efforts to quantify and address the domain gap between synthetic and real-world data for autonomy simulation. By enabling researchers to develop and test autonomous systems in a controlled environment, we hope to bring autonomy simulation one step closer to realization.

Robotics

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the domain gap between using synthetic data and real - world data in autonomous vehicles. Specifically, modern autonomous driving systems require a large amount of testing to ensure their reliability and build trust, but real - world testing has become very challenging due to the lack of large - scale and diverse data sets, especially data sets for edge cases. Therefore, the simulation environment has become an important tool for developing and evaluating these systems. However, existing open - source simulators usually have a significant domain gap between synthetic data and real - world data, which leads to a decline in mobile performance and platform reliability when using simulation data.

To solve this problem, the paper proposes the Scoping Autonomous Vehicle Simulation (SAVeS) platform, which aims to quantify the performance gap between the synthetic environment and the real - world environment and reduce this gap through the introduction of the SAVeS+ extension. SAVeS+ improves the performance of synthetic data in the real - world domain by integrating two domain adaptation techniques, style transfer and depth mask, thereby reducing the dependence on expensive professional sensors (such as LiDAR and infrared sensors) and avoiding cumbersome calibration and synchronization procedures.

Through this series of methods, the paper hopes to develop and test autonomous driving systems in a controlled environment, thereby promoting the development of autonomous driving simulation technology and making it closer to practical applications.