Autonomous Racing
The autonomous vehicle sector has been around for decades, with many academic institutes and established companies dabbling or implementing full solutions. Technological advancements in electronics, fabrication and software have allowed hobbyists access to almost the same technologies used in commercial deployments. Moreover, access to the same concepts can enable hobbyists to understand and implement autonomous vehicles for experimentation and friendly competition.
Beginnings
One example of the earliest community based autonomous competition began in 2009. Sparkfun started the Autonomous Vehicle Competition (https://avc.sparkfun.com/), now retired after ten seasons. The community put together elaborate vehicles based on parts available through online stores accessible by everyone (SparkFun’s website, for example). The AVC helped demonstrate how DIYers/hobbyists can leverage accessible parts to build on the same concepts used in commercial autonomous vehicles. An impressive portfolio of the work done by the hobbyist/DIY community is on display at the AVC website.
Growth
While the electronics/hardware DIY crowd was growing from 2009 onwards, another revolution was taking place. On the software side, the field of artificial intelligence and machine learning (AI/ML) combined with Computer Vision was quickly advancing from academic labs to DIY enthusiasts, most notably the OpenCV project. OpenCV immensely helped the community by providing easy-to-use software development kits that acquire imagery through a web camera and perform various algorithms to gain insight from that imagery. Then, depending on the context of the application, the insight gained from the imagery could be used to provide intelligent feedback to the software systems. The greatest example is autonomous vehicles. Elon Musk has pointed out that the usage of computer vision is highly vital with his Tesla solution (https://techcrunch.com/2021/06/21/tesla-backs-vision-only-approach-to-autonomy-using-powerful-supercomputer/). [Side note: the mention of OpenCV here does not imply its use at Tesla, just an indication that the concept of computer vision is being employed there]
The convergence of the DIY electronics/hardware crowd and the access to computer vision and other AI/ML can be seen in the birth of the AI racing leagues. Two well known AI racing leagues are the AWS DeepRacer League (https://aws.amazon.com/deepracer/league/) and NVIDIA’s DIY Autonomous Car Race (https://developer.nvidia.com/embedded/diy-ai-race). The AWS DeepRacer League supplies their own stock built racer for competitors to program and use in races. In contrast, NVIDIA’s league offers competitors the opportunity to design their own vehicles was some constraints. Both are great ways for the DIY community to get into the fundamentals of building AI racers based on their experience levels.
Future
The skills used in developing AI racers from the basic concepts derived for the AI racing leagues can be seen in motorsports racing. The Roborace series (https://roborace.com/), now in the second season, has demonstrated the applications of those concepts applied to more modern-day applications. The scale-up and modern-day applications do have their own challenges, however. Optimal battery efficiency and computational algorithms designed to address high-speed conditions are just 2 examples of challenges the teams have to deal with. The silver lining, motorsports represents the extreme factors vehicles operate. Applications of lessons learned from motorsport can be passed onto more consumer applications like the basic automobile. The autonomous racing field is constantly evolving, with better technologies being introduced rapidly.
Coming up
Follow this account for more inside looks, updates and news on autonomous racing.
