Formula 1 Aussie legend Daniel Ricciardo is spending his time off the F1 grid by learning about AI as part of Amazon’s DeepRacer League.
During May, developers from around the world can test their ML skills against F1 professionals in the F1 ProAm DeepRacer event, via the AWS DeepRacer console.
The F1 professionals include Dan Ricciardo (recently signed to McLaren for 2021 and beyond), Tatiana Calderon (F1 Test driver) and Rob Smedley (Chief Technical Engineer – F1 Performance, Engineering and Analysis).
AWS Deep Racer allows developers to get hands-on with Machine Learning through a cloud-based 3D racing simulator. Basically, the goal here is to drive a car around a track as fast as possible.
The May race of the 2020 AWS DeepRacer Virtual Circuit will take place on a replica of the official F1 Spanish Grand Prix track. On the Circuit de Barcelona-Catalunya, developers will face off against F1 professionals including the current world record holder for the Circuit de Barcelona-Catalunya, Daniel Ricciardo!
The car is driven not by remote control, but fully autonomously, relying on AI to navigate the course and obstacles.
Programming a car to drive can not be a series of if, this, else conditions, as the sheer volume of environmental variables is too large. Instead an AI is used, with a reinforcement learning (RL) model developed and trained to act like a brain, understanding, and responding to the vision coming from the on-board camera.
This latest video (week 2) is really interesting, as it does a great job of showing how having the right hardware suite (multiple cameras to derive depth and distance of upcoming objects) can help avoid them.
It looks like Dan Ricciardo is having lots of fun with this and it’s a pretty great way to be spending his downtime from Formula 1 racing. Having more developers understand AI, particularly in relation to autonomous vehicles will ultimately help move the industry forward. While its a cute little car you’re racing and Dan’s trademark big smile is out, this really is serious business.
The simulator shows a virtual representation of the DeepRacer, there is actually a 1/18th scale race car (A349.00), that can take the reinforcement learning (RL) models and race around a real physical track.
The competition aspect is essentially a leader board and the best (fastest) AI wins. There are three separate leader boards for each of the tracks – Time trial, Object avoidance, and Head-to-head.