Articles
Arguably one of the most important sensors in collating environmental data is lidar. The laser technology can build a realtime picture of a vehicle’s surroundings and distances between objects, and it isn’t as sensitive to changing conditions that can limit the ability of cameras and radar.
While Elon Musk and Tesla may call lidar an expensive irrelevance on the journey to autonomous vehicles, firms including Ford, Volvo and many others continue to invest in the technology as part of a toolkit needed to make self-driving cars a reality on public roads.
But data stored on research and development vehicles or in the IT infrastructure of OEMs will only deliver limited improvements. If autonomous technology is to progress, data needs to be analysed and used by more people. Which is why Volvo and its lidar supplier partner Luminar Technologies’ decision to share data is important.
Long-range data released
The two firms released a free, curated dataset called Cirrus containing long-range lidar data from Volvo’s fleet of test vehicles. The curated dataset is open to industry developers and researchers in the hope that it will help to bring further advances for safe self-driving.
The raw data from Luminar’s lidar sensors can detect objects ahead of the car up to 250m away and marks a shift in the data and information available to developers. Previous publicly available information has relied on shorter-range, lower-resolution point clouds and uniform scanning patterns, so, while it’s been interesting, its usefulness has been debatable.
Where Luminar’s systems change this is the ability to dynamically adjust scan patterns to optimise resolution up to five times the standard uniform distribution. This is reflected in the dataset, which also includes uniform lidar data and corresponding camera images, and annotations for reference.
The dataset has been released to help advance research and refinement of self-driving software algorithms, in the interest of improved vehicle safety at highway speeds and in complex environments. Duke University’s Pratt School of Engineering in North Carolina is also part of the project.
“Making this substantial dataset publicly available with Volvo Cars represents another significant step towards the development of safe and ubiquitous autonomous vehicles,” said Christoph Schroeder, vice-president of software at Luminar. “Luminar and Volvo Cars are aligned in the belief that sharing knowledge and research will contribute to safer roads for everyone. Our dataset uses non-uniform Gaussian scanning patterns, giving developers extremely high-quality information to help build more advanced autonomous capabilities.”
While the initiative may appear to be an admission that developing autonomous vehicles is more complex than some would have you believe, it’s a smart move. Vehicle manufacturers are facing numerous challenges at the moment, including reducing emissions and the shift to electric powertrains. Using external resources to help develop software algorithms that aren’t part of an OEM’s core expertise takes the pressure off.
Open policy
The lidar data is just part of a move by Volvo to open up the information it holds, allowing third parties to be part of the development process. The Cirrus dataset is part of the new Volvo Cars Innovation Portal, which makes a broad variety of resources and tools available for free, allowing external developers to create new services and in-car apps.
“By making these resources publicly available, we support developers in and outside our company, and collaborate with the best of the best,” said Henrik Green, chief technology officer at Volvo Cars.
Want the best engineering stories delivered straight to your inbox? The Professional Engineering newsletter gives you vital updates on the most cutting-edge engineering and exciting new job opportunities. To sign up, click here.
Content published by Professional Engineering does not necessarily represent the views of the Institution of Mechanical Engineers.