Blog: Driverless cars, same old problems?

Steve Rose


‘Sorry mate, I wasn’t programmed to see you…’


I was reading a story the other day on The Guardian website about an autonomous vehicle (AV) being tested in San Francisco colliding with a motorcycle. It’s not the first time. BikeSocial knows of at least two other incidents where an autonomously driven car has hit a bike. Both of the previous ones involved a Tesla driven on autopilot failing to recognise a bike in front and rear-ending the rider. While Tesla doesn’t claim that its ‘Autopilot’ system removes the need for driver engagement, it seems that some drivers aren’t getting the message. Following the first motorcycle incident in Norway, German newspaper Der Spiegel printed a report from the Federal Highway Research Institute, which had conducted 1000s of kilometres testing on the system and concluded that it represented a ‘significant traffic hazard’ and was ‘not designed for complex traffic situations’.

Interestingly, the incident in San Francisco (which involved a prototype autonomous Chevrolet Bolt, not a Tesla) was just such a complex situation, but one that is entirely familiar to any motorcycle rider. On a three-lane highway, the car, travelling in the middle lane at very low speed (about 12mph) in heavy traffic, detected a gap in the left hand lane and went to move into it. As it did this, a motorcyclist, Oscar Nilsson was filtering (legal in California) through the slow-moving traffic and moved into the gap in the centre lane vacated by the Chevrolet. But before the Bolt completed its move, the truck ahead of it slowed down and the car ‘decided’ to move back into the centre lane. In doing so it’s radar detectors failed to recognise the motorcycle and drove into it, causing Oscar to fall off. There seems to be something about the radar used in the latest AV systems that struggles to recognise motorcycles, which doesn’t bode well for pedestrians or cyclists. 

But it’s not just the radar. In this case the autonomous car caused exactly the same kind of accident that happens to riders around the world every day. And that’s the problem here. I’m not a luddite and we, as riders should be embracing the idea of AVs because, once developed they will make our lives much, much safer. But for that to happen, the companies developing the technology have to start understanding how traffic works for those of us using vehicles other than cars on the road. If you ride a motorcycle or a push bike in traffic, your speed, trajectory and entire journey bears little relation to that taken by a car. We use space in different ways, use speed in different ways and what we do is more akin to complex three-dimensional dynamic trigonometry – judging and predicting how the time and space around will look in three seconds’ time – than sitting in a line of cars, all doing exactly the same thing as the one at the front of the queue.

One of the key comments in the San Francisco story was a spokesman for General Motors (which owns Chevrolet and Cruise Automation, the company trialling the system). He said, “We test our self-driving cars in challenging and unpredictable environments because by doing so we will get better, safer technology on the roads sooner. In this case, the motorcyclist merged into our lane before it was safe to do so.”

I’m sure we all agree with the first part, and it should be noted that the car did have a person in it, supposedly there to rescue a dangerous situation should something go wrong, who failed to act in time.

It’s the second bit that worries me. ‘The motorcyclist merged into ‘OUR’ lane before it was safe to do so.’ That’s the bit that shows the complete lack of understanding about how motorcycles work in traffic. It’s not ‘your’ lane or ‘mine’, simply a piece of tarmac that you used to be in and are no longer. If you move out of it, don’t expect to be able to just go back, without looking if your algorithm changes its mind because something smaller, faster and more nimble might have moved into that space.

The people developing AVs need to understand this alternate dynamic of motorcycle travel. Maybe they need to experience what heavy traffic looks like from the seat of a bike so they can develop their algorithms accordingly. There’s a lot of competition (and potential reward) to get AVs operational and, once perfected, they will make our roads (including motorcycling) much safer.

So, at BikeSocial we’d like to offer an invitation to Cruise Automation, General Motors and anyone else developing AV technology. Let’s meet up and discuss how motorcycles work in traffic. We’ll be happy to take you and any data gathering equipment you have on a pillion ride through some typical two-wheeled commuting situations in either the UK or your own countries so you can tweak your algorithms for a commercial advantage by not using motorcycles as bump-stops.

Please get in touch at and we’d be happy to help all we can.


You have successfully signed up.