All about autonomy, Featured, Self-driving cars, Testing and Simulation

Autonomous vehicle testing could learn a lot from the world of motorsport

By Sonya,

October 15th, 2018

Mark Preston, Team Principal of the DS Techeetah Formula E Team and founder of StreetDrone, believes many self-driving companies are sacrificing safety in the race to deliver an autonomous car. They should be focusing on more rigorous testing.

 

“What has motorsport ever done for me?” is a complaint frequently levelled at top flight racing teams. With the sport’s single-minded pursuit of saving hundredths of a second per lap at huge financial cost, there’s a common misconception that the world of motor racing has next to no impact on road cars.

But without motorsport, the development of everyday innovations such as disc brakes, ABS and traction control would have taken far longer to reach the street. Even rear view mirrors were pioneered in motorsport way back in the 1900s, when race teams realised they could be used to warn a driver of competition behind.

Fast forward to 2018 and we’re in a different race altogether. Waymo has been racking up on-road testing recently as it prepares to launch the first autonomous vehicle. What form that will take is still up for debate, but the former Google self-drive car project hit eight million miles last month leaving rivals in its wake.

The impressive part of Waymo’s accelerated programme is that it is literally streets ahead in terms of test miles driven before a human driver needs to take control. In data compiled by Bloomberg* charting mileage in California in 2017, Waymo’s vehicles travelled on average more than 5,000 miles before human intervention was required. Other companies, such as General Motors, managed just over 1,000 miles, Australian start-up Zoox travelled less than 500 miles, while the rest (Nissan, Delphi, Bosch) covered barely 50 miles.

The fact that human intervention is required at all shows that a completely driverless vehicle is still some way off. Waymo is so far ahead though, that my concern is those behind will take too many risks to catch up. I believe that self-driving companies could learn a lot from the testing that underpins top level motorsport. And I’m speaking as someone who has, I believe, a unique point of view.

Throughout this season’s racing in Formula E, my fears have ranged from not finishing races due to reliability, to making a technical mistake with the car or our drivers getting caught up in events that are out of their control on the track.

But at the same time, we have had to push hard. We’re racing after all. As a relatively small privateer team, Techeetah has been competing against large automotive companies such as Audi, Renault and Mahindra. However, there’s a big difference between pushing on the race track and the public road. Within the world of autonomous vehicles there have been several high-profile incidents, which have showed that tech companies may have been pushing too hard, forgoing rigorous testing and process for speed of development.

When I started out in racing, we had so many unknown unknowns. We didn’t have data loggers in Formula Ford, so I had to write my own simulation software and by building digital models, I learnt a lot about how we correlate to the real world.

When I moved to F1 at Arrows and became Head of R&D, many of the innovations I put into place were tools, such as software and hardware test rigs, that allowed us to more rigorously test new ideas and components before we took them to the test track.

Mechanical parts were – and still are – designed and tested using FEA software, fatigue tested on rigs in a laboratory and only signed off after extensive testing on a track. That’s all before going to a race. There is always a process and risks are minimised. And that’s a key point. Even when racing teams are taking risks, racing cars are still caged animals because testing, and racing, takes place in a controlled environment: the race track.

Some Autonomous Vehicle companies are taking needless risks with their approach to testing on the public road. They shouldn’t be experimenting with fundamental parameters on the public road. Validations like these must happen in simulators and on test tracks before the car ever reaches the public road.

Part of the reason we built our own autonomous vehicle was because we were considering entering the Roborace championship and saw that each race car was quite expensive. What we wanted was a low-cost way to repeatedly experiment, hence the StreetDrone ONE (our Renault Twizy-based self-driving car).

A fundamental idea behind StreetDrone has been to have a simple test platform that’s priced low enough to allow many more real-world experiments to happen. But backed up by a robust simulation process. Why? Because there are currently too many unknown unknowns. To speed up development in safety, we provide elements learned from motor racing such as hardware-in-the-loop systems which can also include a full car immersed in a simulator.  The correlation between each level of simulation and the track is paramount before the software finally makes it to the road to be tested with a safety driver onboard.

Tech companies need to realise that there is often no substitute for good old fashioned engineering to figure out a problem. In the world of racing, the rate of development is not slowed down by digging deeper. More often than not, testing will teach us something new and we will go faster. Motorsport has done a lot to improve road car safety and it can do the same for autonomous vehicles too.

Become a Member

Enter our draw for a chance to test drive the StreetDrone ONE and spend a day with the SD team! Additionally, get acces to exclusive technical content, discounted events and conferences