One woman killed by an Autonomous Car in a recent incident
Local police in Tempe Arizona has confirmed that a woman pedestrian was killed on Sunday by a self-driving car which failed to react to her crossing the road in front of it. Although a safety driver was behind the wheel, the vehicle was in self-driving mode, and neither the driver nor the vehicle reacted in time to stop a collision, resulting in the fatality.
This was not the first accident involving an autonomous vehicle. A previous incident, in May 2016 involved a semi-autonomous Tesla electric vehicle. The occupant, officially the driver, was killed at an intersection, in a crash involving a truck.
The Tesla accident and the memory of the incident 2 years ago will almost certainly delay the launch of driverless taxi services, which were planned to commence as early as later this year.
Autonomous vs human drivers – accidents per mile
Judged by the two fatalities seen so far, autonomous cars have a poor driving record compared to their human counterparts. According to the Association for Safe International Road Travel:
- 3000 people per day, 1.5m per year, are killed every day, worldwide, on the roads.
- Another 20 – 50 million are injured annually, many seriously.
- In the USA, where the incident occurred, over 100 people a day are killed
- There is one road death per 100 million miles driven in the USA.
Waymo, the Google / Alphabet subsidiary which has covered the most distance in tests for a fully autonomous vehicle release, has travelled only 5 million miles, in total, on public roads. Uber, which has also invested heavily in self-driving car technology, has covered less than half that distance, currently only approximately 2 million miles.
Is it too early for entirely self-driving cars?
The focus of current self-driving vehicle technology as it stands is to have them act entirely autonomously. Industry analysts and the groups which represent them point out the advantages of connecting self-driving vehicles to the mobile internet, to improve their safety performance.
Existing 4G connectivity data transmission speeds are insufficient to transact the volume of data points collected by autonomous cars. However, 5G, the new generation of cellular networks which could transmit such large quantities of data, is on the sidelines and scheduled for release over the next 2 years. Should regulators insist on autonomous car testing which involves this technology to minimize the number of accidents?
5G Americas, a consortium of cellular network ecosystem members suggests that 5G (and an associated technology, the Internet Of Things) will improve the safety, speed and intelligence of driverless cars. V2X – Vehicle to Everything – technology, will, they say, allow autonomous vehicles to interact with each other and street features like traffic lights, to improve the information they are using to make their decisions. Additionally, once connected, cars could share the data they have immediately, making it easier to collate and for the ML (Machine Learning) algorithms operating behind the scenes to ‘learn’ safer behavior faster.
Should self-driving cars have to undertake driving tests?
This most recent fatality, although deeply regrettable, does not appear to have impacted public sentiment on the value of autonomous cars. Neither is it likely to. The most probable outcome is that the matter will be dealt with out of court, with Uber ‘keen to settle in order to avoid a test case’, reports the MIT Technology Review.
The death does, however, raise natural questions about the attentiveness of the safety drivers which support and are responsible for the vehicles involved in Autonomous car tests – perhaps understandably if they are driving 8 hours a day with nothing to do but watch the car drive itself. It also raises the question of the ‘right’ level of testing and the ‘right’ types of technology to use, for regulators.
Self-driving cars should have to undertake driving tests before being allowed to roam the streets, just as humans do. Governments should not necessarily insist on safety standards for those driving tests which are perfect (or even necessarily close to it), just some which are better than the current high but accepted road accident toll.
Autonomous cars will certainly become a feature of our roads in the months and years ahead. However, given that current statistics indicate self-driving cars are substantially below existing safety levels, it seems prudent to insist on an upgrade to the safety mechanisms and technology employed in driverless cars. If that means we need to wait for internet connectivity (including, potentially, 5G) to be available, so driverless cars can interact with each other and their environment, rather than by tacitly accepting an individual view of it, the delay would seem to be worth it both from a human impact, and a public perception perspective.