Autonomous Vehicles


A blog about Self-driving cars, Deep Learning, Computer Vision and Ethics



This blog presents the results of a short technological watch about Autonomous Vehicles as part of an exercise of the École Centrale de Lyon Computer Science and Data Science Masters degree. Its goal is to present the state of AV as of March 2020.

Autonomous Vehicles, now



Since 2018, the industry have been developing really fast and is now in the middle of a fast pace competition. However, as the technologies and the prototypes moved forward, the buzz words in the media switched from “robot-taxis”, “self-driving cars” to “safety” and “advanced driver assistance systems” (i.e. semi-autonomous vehicles). The reason ? The fall of the hype around AV and a global disillusionment following the death of a pedestrian in Tempe, Arizona, during the test of an Uber self driving vehicle.

2020 is set to be the year of strategic alliances, safety-focused development and slow deployment or tests of semi-autonomous vehicles to fully-autonomous vehicles, in a very controlled space. While the corporate capitals and ventures seem infinite, the main goal for AV manufacturers is now to gain people’s confidence. A study by Gartner says that the technology could be mature by 2025.

Nevertheless, autonomous vehicles are a very important subject for cities and governments all around the globe, that aim to make the roads safer and to improve the population quality of life. Many major projects are being carried out in developed countries, such as the UK, Germany or Japan :


Some important projects carried out by governments :





2020 also shows the diversification of Autonomous vehicles :





Still a promising field :




References
Challenges of 2020 by techcrunch.com (January 2020)

Innovation in Technology



2020 is no exception to the recent depth rush for neural networks. GPUs have become faster, cheaper and above everything, more powerful. NVIDIA now provides the best and afforadble components the world has ever seen, allowing the Autonomous Vehicles systems to perform better than ever. Research in the field of Deep Neural Networks and Computer Vision have also received a lot of funding as many companies rely on it. Thus new architectures and techniques have driven the performance even further.

Therefore, an important part of innovation comes from Deep Learning service provider, that are turning research into active products. AI startups such as China's first algorithm provider SenseTime (most valuable Unicorn to date with a $4.5bn valuation) for example, are innovating and creating partnerships with car manufacturers that are willing to use their state-of-the-art algorithms at their benefits.

IoT and sensors are also a field of research and active innovation. Companies like Quanergy or Swift Navigation have developed very powerful laser sensors and GPS chips that allows Autonomous Vehicles to perform better. With better data comes improved performance. The same goes for hardware as shown in Tesla's tweet below.


What the autopilot sees :






Successful AI companies' partnerships:



(old tweet)


2020 also see the development of self-driving test cities :
After 3 years of testing in real cities, it seems now safer for some countries to build fake cities.



Main Players



The main autonomous vehicles players in 2020 are composed of the well known self-driving companies, a few newcomers and the famous ancestral car manufacturers that joins the run to semi-autonomous vehicles.




(former Google project)





















(from General Motors)







Cadillac (like other famous car companies)






Small players example: Navya (French group)



Ethics



In addition to the classic philosophical issues that have been widely discussed (what the car would do in lose-lose situation like killing 2 people by going to the left or killing 1 person by going to the right), 2020 is set to be the year of many ethical questions regarding self-driving cars.

The first main subject is security. After the death of a citizen in Tempe, Arizona as part of Uber self-driving car testing, the first ethical question concerns the government's decision to allow public testing when the systems are obviously still work in progress. A second matter is cybersecurity: is it okay to put self-driving cars on the public roads, when they can be easily hacked ? Like every computer system, it can't stand without flaws and breaches. Some scientists of the Ben-Gurion University of the Negev in Israël have even proven in early 2020 that Tesla's autopilot could be easily tricked.

Then come to mind the subject of privacy. As technology comes into the car, ad-tech companies see an opportunity to drive benefits from it with e-marketing. With all the sensors and algorithms contained in self-driving cars, people's vehicles will become huge data gathering machines. What about privacy and consent ?


Cybersecurity threats :







Privacy concerns :




(old tweet)

Lancelot Prégniard - MOS 4.4 "Nouvelles Technologies de l'Information et de la Communication" 2020