What direction now for autonomous cars?
The recent road traffic accident in Arizona, in which a cyclist was fatally injured by a self-driving car undergoing road tests is a personal tragedy for the family involved, but also calls into question issues about the way these vehicles are being tested in public.
The incident happened in Tempe, Arizona, one of the first cities in the US to approve testing of self-driving vehicles on public roads, with the proviso that a safety-driver was present at all times who could intervene to avoid accidents. Uber is one of the companies developing self-driving vehicles and it was its test vehicle, a large Volvo SUV, which was involved in the evening accident. The cyclist, a 49 year old woman, was pushing her cycle across the highway (not cycling), when she was hit by the Uber travelling at approximately 40mph, and later died in hospital.
This first fatality from self-driving vehicles obviously created a media stir, and shortly afterwards, Tempe Police Chief Sylvia Moir issued a statement in which she said Uber was likely not at fault, that the vehicle was not speeding, that the cyclist should have crossed further along the road, at the crosswalk (traffic lights), and that the cyclist was wearing dark clothing and no lights. Moir also reported that the safety driver had said "it was like a flash, the person walked out in front of them, his first alert to the collision was the sound of the collision".
Regardless of how much fault lay with the pedestrian, surely we have been told many times that autonomous vehicles are so packed with safety features that they would have detected and reacted to the pedestrian far faster than any human driver could have. The weather was good, and the Uber vehicle is equipped with radar, lidar, and cameras to detect exactly this sort of event. Although the police were quick to say Uber was not to blame, they also said that the vehicle showed no signs at all of stopping. Any speed reduction would have increased the survivability of collision considerably.
Tempe police later released dashcam footage of the seconds before the accident. The news stations, having seen this, tended to reinforce the view of the police chief. The ABC news reporter said the video "does show how difficult to see the victim on that dark road", the CBS reporter said "The pedestrian is only visible for about a second before the crash" and that "the accident may have been unavoidable". Various other commentators have used phrases like "comes out of nowhere".
It seemed cut and dried, if a pedestrian steps into the path of a car, no matter how smart the technology, the laws of physics says there is a point where you simply cannot stop in time. Yes, it seemed cut and dried,... until I decided to watch the dashcam footage. It wasn't easy to watch, knowing the outcome, but the video stops immediately prior to the moment of impact, and watching it completely changes my view of the incident.
Yes, it does look like the pedestrian comes out of nowhere, but being a dashcam user myself, the first thing I noticed about this footage is that it is way too dark. I know that the human eye sees much better than the dashcam at night, but even so, this is deceptively dark footage. A search of youtube reveals the following dashcam footage of exactly the same section of highway, shot by another concerned motorist a few evenings later.
This footage shows Mill Avenue is not an unlit road in the middle of the desert. It is in fact a well-lit urban road, with gentle curves, and at 40mph you can easily see features for ten seconds before you reach them. Seeing that video you realise the pedestrian was three quarters of the way across the road at impact, and the idea that they "shot out of nowhere" and were "only visible for a second" just doesn't hold water. It is disappointing that the major news services, who work with video day in and day out and know its capabilities, have not practised some investigative journalism of the Uber footage.
The second very disturbing fact to emerge from this footage is the interior view of the "safety driver". Critics of public testing of vehicles have often pointed out that safety drivers sitting in the driver's seat but doing nothing except look out of the windscreen will find it difficult to keep alert enough to intervene in a potential accident, but no-one expected the safety driver to be spending so much time with her eyes off the road entirely and apparently looking down at her phone.
The motor industry and technology companies have been rolling out the PR machine and wheeling out automotive experts, pointing out that the various prototypes have covered thousands upon thousands of miles and this is the first fatality, which is already much safer than human drivers, and that the only way to iron out the bugs is by testing on real roads and motorways, (something which the UK government is eager to allow). That may be true, but pedestrians did not sign up to be test crash dummies for the auto industry.
28th March 2018
This article comes from the SKILLZONE email newsletter, published monthly since January 2008, and covering topics related to technology and the internet. All articles and artwork in the SKILLZONE newsletter are orignal content.