GTC Europe Live Blog: What is Level 5 Autonomous Motorsport?
by Ian Cutress on October 10, 2017 9:59 AM EST- Posted in
- AI
- Trade Shows
- NVIDIA
- Live Blog
- Automotive
- Autonomous
- GTC
10:01AM EDT - One of the talks at GTC EU today is titled 'What is Level 5 Autonomous Motorsport?'. Live blogging time!
10:01AM EDT - Bryn Balcome, part of Roborace on the stage
10:01AM EDT - Level 5 Autonomous Motorsport is coming
10:01AM EDT - >The speaker has a history in F1 and motorsport
10:02AM EDT - Going through some racing history
10:03AM EDT - Autonomous Motorsport was predicted years ago
10:03AM EDT - OEMs are very interested in power. A picture of a Bugatti Chiron on the screen
10:03AM EDT - Now we talk about cognitive power, not mechanical power
10:04AM EDT - Demonstrating cognitive intelligence in the crucible of motorsport
10:04AM EDT - 'Motorsport will split between pure racing and technology'
10:04AM EDT - CEO on an info video
10:05AM EDT - 'Creating an automotive motorsport to accelerate consumer delivery of this technology'
10:05AM EDT - Oh ok, the CEO is Lucas Di Grassi
10:05AM EDT - Robocar on a track at Paris during the Formula E weekend
10:06AM EDT - One year from concept to the car being designed
10:06AM EDT - Racing drivers have emotions and characters that take time to develop - AI can be trained to have these emotions
10:07AM EDT - 'Incidents such as F1 big crashes shouldn't happen' - wait what?
10:07AM EDT - Formula 1 is becoming less relevant to road racing each season and going in a different direction
10:08AM EDT - Le Mans style racing is even more different
10:08AM EDT - F1 and Le Mans doesn't develop ADAS systems to assist the drivers
10:08AM EDT - Arguably real road environments are more difficult than racing
10:09AM EDT - Racing is about car control, which can be developed through motorsport
10:10AM EDT - Audi and BMW have both showed racing around a track using autonomous systems, including drifting at 'high' speeds
10:10AM EDT - Optimal car control is not a solved problem, but will get there
10:10AM EDT - Now it's about situational awareness
10:10AM EDT - How cars react to different scenarios
10:10AM EDT - Such as avoidance of new obstacles of an environment it already knows
10:11AM EDT - Racing is all about interacting with other drivers and implementing your own influence in an environment
10:12AM EDT - 'Think about Pikes Peak with traffic, and imagine the challenge for an autonomous vehicle racing side by side through traffic'
10:12AM EDT - It is critical that your driver is intelligent, not just raw pace
10:13AM EDT - Bringing research to life so the public engages
10:14AM EDT - Street racing helps develop road relevant AI capabilities
10:15AM EDT - Develping AI to assist drivers - using AI between spotters and drivers in Nascar
10:15AM EDT - Already used in Rallying and side-car racing
10:16AM EDT - (Rallying and side-car use two drivers that collaborate - AI should be the same in other motorsport and on the road)
10:16AM EDT - e.g. the AI takes over in critical scenarios if the driver does not react to an obstacle
10:17AM EDT - Human drivers in autonomous simulators to help train the AI with suggested movements
10:18AM EDT - Q&A time
10:19AM EDT - Q: Timeline to develop autonomous motorsport
10:19AM EDT - A: We'll see small step progression and competition formats evolving to set new targets and challenges. Evolve as the industry evolves
10:20AM EDT - Q: How many contenders does Roborace have?
10:21AM EDT - A: We only need two cars to compete. Over 150 companies have showed interest to be involved (although these are OEM). We're focusing on PhD/Postdoc level competitiveness
10:21AM EDT - Q: Is the hardware fixed ?
10:22AM EDT - A: The hardware vehicle element is one, the intelligence platform is another, and a third is then AI drivers. This competition wants to focus on the AI drivers segment
10:25AM EDT - Q: How are you going to show the public that these AIs are more than ones or zeros, where real world drivers have emotions?
10:29AM EDT - A: It's all about interpreting risk. When we watch a driver in the car, we interpret their actions and how they manage risk. AIs will be similar - the management of risk will lead to developing personalities that will play out on the track
Facebook
Twitter
RSS
7 Comments
View All Comments
BedfordTim - Tuesday, October 10, 2017 - link
This sounds a great way to diminish public confidence in AI drivers. Racing is about pushing the limits and the almost inevitably means the odd crash, which is that last thing you want to see an autonomous car do.Wardrop - Tuesday, October 10, 2017 - link
What if they don't crash. How confidence inspiring would that be? A driver with 360 degree vision, almost perfect physical prediction, etc. I think it'd be hard for them to crash, and I assume not crashing is probably the primary objective the AI driver, after which is driving quickly. Crashing costs money lets not forget.zodiacfml - Sunday, October 15, 2017 - link
Not money but safety. A crash means a failure for AI. They might set the rules with costlier penalties for crashes, thus, algorithms will be set to prioritize safety/control than being ahead.0iron - Wednesday, October 11, 2017 - link
Racing in a circuit is in controlled environment. Real life traffic is much more complex. Driving during power cut-off in a city where traffic light is not functioning, flooding, landslide or any disaster will be difficult for AI.Valantar - Wednesday, October 11, 2017 - link
I get fixing the vehicle hardware, but fixing the AI platform seems arbitrarily monopolistic. Shouldn't any company claiming to have the hardware and/or software chops to compete be able to do so? Are Nvidia scared of being beaten at their own game? Can any relevant argument to fix the AI hardware platform actually be raised?charleswells889 - Friday, November 3, 2017 - link
Thank you for sharing. https://ios12update.com/harryseth - Wednesday, April 25, 2018 - link
This sounds like a big deal to me. I missed the live updates, but reading them now!