Aurora Innovation sets the stage for advanced research in autonomous vehicle localization by unveiling the Aurora Multi-Sensor Dataset. This is a leap forward for researchers and developers aiming to harness the power of diverse weather and traffic data.
Aurora Innovation, Inc., in collaboration with the University of Toronto, announced the release of its Aurora Multi-Sensor Dataset. This sizeable open-source dataset, boasting advanced localization ground truth, was initially presented as PIT30M at the renowned International Conference on Intelligent Robots and Systems (IROS) 2020.
Why It Matters
The Aurora Multi-Sensor Dataset is a revolutionary contribution to the world of autonomous systems research. It offers rich metadata spanning various weather patterns across the four seasons, different times of day, and a multitude of traffic conditions. Its scope is considerably broader than other publicly accessible localization datasets, making it an invaluable resource for developing and evaluating large-scale autonomous vehicle localization strategies.
Presented previously as PIT30M, the dataset gained recognition at the International Conference on Intelligent Robots and Systems (IROS) 2020 as a Finalist for Best Application Paper. The paper, titled “Pit30M: A benchmark for global localization in the age of self-driving cars,” was authored by an interdisciplinary team of researchers. Aurora’s intention in disseminating this data to the wider academic community is to foster engineering research that will further advance the autonomous systems field. The dataset’s size and variety allow its application in diverse research areas like 3D reconstruction, HD map construction, and map compression.
The Aurora Multi-Sensor Dataset, collected between January 2017 and February 2018 in Pittsburgh, PA by Uber Advanced Technologies Group (ATG) – later acquired by Aurora in January 2021 – is now hosted on Amazon Simple Storage Service (S3). The dataset, obtained using a high-resolution LiDAR sensor and seven high-definition cameras, is accessible through the Open Data Sponsorship Program for non-commercial academic use under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License (CC BY-NC 4.0). Aurora Innovation’s initiative in sharing this dataset represents a significant stride toward robust and meaningful advancements in autonomous vehicle technology.