Why safety tech and speed limiters can’t save us all

Have a read of the caveats and disclaimers in any number of current owner’s manuals describing how ADAS works, or doesn’t. Volvo says its City Safety system can detect cyclists, but only adult cyclists riding adult-size bicycles and not “partially obscured cyclists, or cyclists if the background contrast is poor, or cyclists wearing clothing that obscures the body outline, or bicycles loaded with large objects”. 

Pedestrians, too, have to be very pedestrian-shaped, and not “partially obscured, shorter than 80cm or carrying large objects”. And if your car’s cameras and sensors are dirty or obscured, then all bets are off.

It’s certainly not just Volvo. Subaru’s Eyesight driver assist tech works extremely well by the standards of these things, but the circumstances under which it doesn’t are listed at length. The latest Honda CR-V has just scored five stars in the Euro NCAP tests, but the detection limitations of its Collision Mitigation Braking System, Adaptive Cruise and Low-Speed Follow systems are at least as extensive as Volvo’s, perhaps more so. These are just a handful of examples.

When all is working as it should, detecting something isn’t necessarily a problem, but deciding what that thing is and then determining whether it should be avoided is the issue.  

If a pedestrian carrying a ladder or someone riding a bicycle while wearing a big hat is going to confuse the best current ADAS set-ups, imagine how they’d cope with large groups of bicycles going this way and that, or a bunch of people crossing the road together, all wearing and carrying different things – hardly unusual urban scenarios. 

The technology simply isn’t yet able to make these decisions as effectively as a fully operational, fully attentive human. And this is a burning issue for proponents of self-driving cars. 

A few weeks ago analysts at KPMG published what has become its annual Autonomous Vehicles Readiness Index, a global league table of which countries are best prepared, or best preparing, for the arrival of self-driving cars. 

The winner was The Netherlands (the UK came seventh), but even so, the report noted that the country “has a lot of bicycles” and that “in urban areas it will be very difficult to start autonomous driving”. Their conclusion? “It may make more sense to keep transport modes separate rather than integrating AVs to work there.”

And that seems to be the way the thinking is going: if we can’t make our roads completely safe, we’ll have to make them if not completely then certainly considerably different. The best way, perhaps the only way, to avoid pedestrians and cyclists getting killed by cars, regardless of blame, is to keep them apart.