The age-old adage, “Race on Sunday, Sell on Monday,” is being applied to driverless cars in Roborace, a global championship series for autonomous electric race cars. Rather than fender-to-fender duels between race drivers, competitors will be programmers. The ones with the best software and artificial intelligence (AI) techniques will be taking the checkered flag. These advanced technologies could be used in future driverless vehicles that will be sold to consumers in coming years. As expected, the key challenge is collision-avoidance. If a driverless racer can avoid others racing alongside at 200 mph, the technology stands a pretty good chance on the street.
Each of the 10 teams participating in the Roborace series will be competing in identical driverless Robocars, two per team. It’s a new take on spec series racing where teams compete in identical cars, but in this case what sets teams apart is not a driver’s skill and daring, but the algorithms and capabilities of its programmers. In other words, the best computer programming skills will result in a win with less requirement for the enormous budgets or huge R&D required for most race competitions. That means college teams could conceivably compete against a team of Ferrari engineers. Racing is planned for the same tracks used by the FIA Formula E Championship series where electric-powered race cars compete, but still with human drivers.
Designed by Daniel Simon, the 2145 pound, primarily carbon fiber Roborace Robocar is powered by four 402 horsepower (300 kW) motors and a 540 kWh battery, plus the requisite electronic gear. Obviously, there is no need for a cockpit with a steering wheel, instruments, or pedals. Safety equipment like roll cages and air bags are also unneeded. This frees up space and weight in the race car for a huge array of electronics including two radars, five laser-powered LIDAR detectors, six AI-driven cameras, two optical speed sensors, 18 ultrasonic sensors, and GNSS (Global Navigation Satellite System) positioning. The Robocar’s nose is made of special material so radar can ‘see’ through it. LIDARs are built into the wheel arches to eliminate blind spots. Computing is done by an NVIDIA Drive PX2 AI supercomputer capable of up to 24 trillion AI operations per second.
Initial testing began in the summer of 2016 using ‘DevBot’ test vehicles. These had the same internal components including battery, motor, ad electronics used in the Robocar, but were placed in the chassis of an LMP3 Ginetta race car. DevBots drove on their own, but they also had a cockpit so an engineer could sit inside and take control if required. The DevBots were quite different from the Robocar in looks and performance. During testing before the 2017 Buenos Aires ePrix, two DevBot cars raced autonomously against each other for 20 laps. This was the first-ever live demonstration of two driverless cars on the track at the same time. One successfully avoided a dog that ran onto the course while the other car crashed on a corner, showing there were clearly many challenges to be solved before race fans would see a full grid of Robocars racing on a track.
Now another milestone has been achieved. A self-driving Robocar performed a demonstration on the city streets of Formula E’s Paris ePrix on May 20, with the car negotiating its way around 14 turns of the circuit in self-driving autonomous mode. Similar demonstrations will be performed at other Formula E events during the rest of the 2017 racing season.
There remains one important question: Will motorsports fans want to see silent driverless cars racing? One pundit says maybe so…when the NFL uses robot quarterbacks. Still, progress marches on. The Robocar is taking a bold step toward a new type of racing that will provide learnings and technology breakthroughs that should help bring autonomous cars to our highways sooner than later.
