Aardvark Daily

Please visit the sponsor!

Did bad technology kill Elaine Herzberg?

They're all the rage... cars that can navigate and drive themselves, with the person in the drivers' seat simply going along for the ride.

Some Tesla have self-drive and Uber is investing a fortune in self-drive trials around the world.

What surprises me is that this self-drive technology is pretty much unregulated.

That's even more ridiculous when you consider that we're talking about a two tonne vehicle which can cruise the open highway at 100Km/H and must interact with other vehicles, cyclists and pedestrians.

Sadly, something went wrong in Arizona this week and a woman pedestrian, Elaine Herzberg, was killed by one of Uber's self-drive cars.

Yet in virtually every country around the world, there is more regulation associated with a child's lightweight flying toy drone than there is with self-drive cars.

So what went wrong in Tempe Arizona? Why didn't all that fancy technology prevent the Uber self-driving Volvo from killing a woman as she pushed her bike across the road?

Well I don't know that there's any single answer to that very important question. More likely this was the result of a sequence of unfortunate events, none of which on its own would have produced the fatal outcome but, when all happened simultaneously, someone died.

Let's look at the vehicle and its onboard technology first.

This self-drive car is festooned with all a range of sensing technologies that are supposed to stop it from crashing into people, cars, buildings and anything else it's likely to encounter.

There are a number of cameras (the early versions had about 20 cameras but now I think they're using only seven or so). These cameras serve a number of roles such as allowing the vehicle to see the boundaries of the lane it's in, read street-signs and sense the potential for collisions. Most of that camera data is fed straight into the computers in the boot which are responsible for analyising it all and coming up with corrective inputs to the controls.

Then there's the LIDAR -- a ranging laser system which sits atop the roof and provides a 360-degree "view" of the world. If you want to see how LIDAR works and interacts with the onboard computers, I suggest you Google a very interesting TED talk on the subject. Why haven't I included a link here? Hell, I'm not your mother! :-)

Finally, there are anti-collision radars mounted around the vehicle. They operate using microwaves to detect objects moving towards or away from the vehicle and to provide cues for emergency braking in the event that an obstacle or stationary vehicle appears in front of the car.

So we have what amounts to three levels of sensing, any one of which ought to have been able to sense the pedestrian and trigger an emergency braking maneuver. However, looking at the onboard camera footage from the incident, it does not appear that the car braked at all before the impact.

So the technology clearly failed to work well enough.

But what about the supervising driver, why didn't she react in time to avoid the crash?

Well, as seen in the footage above, she appeared to be busy, perhaps using her smartphone and although she looked up a few seconds before the crash, her gaze returned to her lap in the seconds immediately before the impact.

Could she have seen what the self-driving car's technology couldn't?

Well if you look at the footage of her, you'll see through the side window, that there was a reasonable amount of lighting in that area. You can see the side of the road quite clearly in the seconds before impact so I suspect that the forward-facing camera was set up with relatively low gain so as to avoid the area in the headlight beams from "blowing out". This would make the shadows look *much* darker than they really were. It's quite possible that if the supervising driver was actually looking ahead, she may well have seen the pedestrian in time to hit the brakes.

Finally, what about the pedestrian herself?

She appeared to be completely oblivious to the approach of the Uber car. Why?

Surely, crossing *any* road as a pedestrian requires you to look both ways and ascertain that there are no approaching vehicles. Why did she not do that? Was she perhaps wearing earbuds and engrossed in listening to something on her smartphone? Or was her judgment impaired in some way, perhaps due to fatigue, alcohol or some other substance?

From this I hope it is clear that it probably took all of these factors to coincide for this death to occur.

If any one of the car's sensors had triggered the automatic braking, the woman may have either survived or been missed completely.

If the supervising driver had been looking ahead at the right time she'd probably have seen the pedestrian leave the footpath and start to cross the road -- thus prompting her to hit the brakes.

And, if the pedestrian was paying attention, she would have seen the oncoming Uber car and waited for it to pass before attempting the crossing. I shudder to think what the effect of super-quiet EVs will be on the number of pedestrian deaths in future.

So who's to blame?

Well everyone's to blame. The Uber technology developers because clearly their pedestrian avoidance technology is not fit for purpose (yet). The supervising driver because she wasn't paying attention -- which is why she's in the car in the first place). The pedestrian because she had an obligation to ensure that it was safe to cross.

So, from half a world away, I remain utterly confused as to why it is that anyone can get behind the wheel of a car with unproven self-drive technology and put the lives of everyone around them at risk... yet the neighbour's kid can't fly their 25g toy drone without getting a raft of permissions and learning all about airspace classifications first.

Lunacy!

Footnote: Please tell your friends about Aardvark Daily, don't keep this column a secret

Please visit the sponsor!

Have your say in the Aardvark Forums.

PERMALINK to this column

Rank This Aardvark Page

Change Font