Roborace is formulating ideas for its first autonomous racing championship planned for 2021, according to the company’s CEO Lucas di Grassi.
The AI racing concept is currently midway through ‘Season Alpha’, the name given to the first year of head-to-head competition between different teams.
This mainly experimental phase involves academic institutions with which Roborace has forged partnerships, including the University of Pisa and the Technical University of Munich.
It will then embark on Season Beta next year before looking to incorporate the technology and acquired knowledge in a more official competitive setting in 2021.
Exact details of the autonomous racing championship’s format are yet to be decided, with Roborace currently concentrating on how the sensor-led technology can be used.
“The idea is that we have our own competition series in the future, which does not need to be traditional racing,” di Grassi told e-racing365.
“It could be a competition of obstacle avoidance. It could be a pitch-black night race. Some types of software require computer vision [for night time] but with LIDAR you don’t.
“So it will be different challenges and not traditional racing. Traditional racing will be a part of it but will be just a compliment to the other things that we do during the season.”
Di Grassi suggested that Roborace will provide a base car for the 2021 season, in a similar strategy to that employed by the Extreme E electric off-road series.
“In season one, we will launch a proper competition championship where the teams have their own base pack and improve on that or develop their own,” he said.
“Nobody in the world can come and program their software and run it at the moment because it doesn’t exist. So we need to develop a certain level of base code that anyone can do.
“You might want to build a Roborace team. We can give you the base layer and the car will drive itself. You’re then going to fine tune the downforce and corning capabilities.
“We are trying to get a stable and reliable base in these first two years, and then by season one it should be good enough.”
Last year, Roborace decided to introduce its current LMP3-based DevBot 2.0 prototype in place of the original Robocar, as a way of enabling driver/AI interactions.
The move was made partly to emphasize the autonomous nature of the Roborace vehicle system, which is not remote controlled, and di Grassi believes this complimentary formula will be the basis of the championship proper.
He envisions the format developing to the point that AI and human drivers could even share stints in future editions of the 24 Hours of Le Mans.
“I could say anything [about 2021] but I think things will change,” he said.
“The tendency is to be more AI as the future goes on, but I think the human element is important, especially in sports cars: this category only exists because there is the pleasure of human driving.
“We need to have the human element complimenting it and helping it, but I see Roborace doing some races on full AI mode, but the combination is really good, maybe for five to 10 years.”
Di Grassi said Roborace wants to mobilize DevBot 2.0 for a series of public runs at world-class motorsport events in the near future.
“We moved away from Formula E because we had very little track time and it’s the most difficult place to develop,” he explained, referring to Roborace’s early public runs during E-Prix weekends.
“The track is built last-minute in the cities and some of the corners are super tight for our cars, so it’s not the perfect scenario.
“We hope that we can do a multi-platform Roborace series in the future, maybe one event at Goodwood, one at Formula 1, and one at Le Mans.”
Goodwood Timed Run a “Very Important” Event
Di Grassi was pleased with DevBot 2.0’s showing at the Goodwood Festival of Speed last weekend, where the car managed a timed ascent of the famous 1.16-mile hill climb.
Dense tree cover meant the course lacked a reliable GPS reception, which challenged DevBot 2.0 to use LIDAR detection to feel its way up the route.
The car completed the course in a time of 66.96 seconds, which was almost 10 seconds quicker than the Robocar’s unofficial time from last year.
Di Grassi also took part in one demonstration where he drove the car manually up part of the track before stepping out and leaving DevBot 2.0 to complete the ascent solo.
“Goodwood was very important because it’s one of the few places where GPS doesn’t work properly,” said the Audi Formula E driver.
“The car has to use computer vision in general to locate itself, and the track is very narrow.
“There are no clear limits, so the competitor needs to identify what is grass and what is a shadow, what is tarmac, and make its decision.
“It was a good challenge for us, and there were a lot of people, so it was a very important event.”