Uber’s recent tragic driverless car accident that saw a woman lose her life exposes the risks and challenges of self-driving cars technology.
The past week saw one of Uber’s self-driving vehicles hit a female resident of Tempe, Arizona, who later passed away due to her injuries at a nearby hospital.
While this is not the first time an assisted-driving vehicle is involved in a fatal accident (some may recall the Tesla Model S with Autopilot engaged that crashed into a trailer back in May 2016), it is the first time a vehicle labelled as supposedly ”autonomous”, or self-driving does so.
Arizona has long been a haven for the testing of self-driving cars, both due to the state lenient regulations as well as its favourable weather conditions. This incident comes just weeks after Arizona Governor Doug Ducey revised an executive order to allow fully driverless cars on the state roads.
Safety questions raised
The accident happened on Sunday evening when the Uber Volvo XC90 equipped with self-driving technology travelling at a speed of 40mph (64km/h) struck the 49-year old woman. At that time, 44-year-old Uber test driver, Rafael Vasquez, was behind the wheel and the car autonomous mode was engaged.
The victim, Elaine Herzberg, was walking her bike along the road. As the self-driving Volvo XC90 approached, Elaine decided to cross the street without using the crosswalk. The crash then followed as she stepped out in front of the vehicle.
The accident is still under investigation by the Tempe Police Department who have released a video showing the fatal collision from the vehicle’s cameras.
Tempe Police Vehicular Crimes Unit is actively investigating
the details of this incident that occurred on March 18th. We will provide updated information regarding the investigation once it is available. pic.twitter.com/2dVP72TziQ
— Tempe Police (@TempePolice) March 21, 2018
The video shows that the accident seems to imply it was unavoidable, as neither the car nor the driver was able to anticipate the woman stepping out into the road.
Uber has nonetheless grounded its fleet of self-driving vehicles, halting testing at all its four North American locations. Regardless of what the Tempe investigation findings, there are some very real and pressing issues related to self-driving cars.
On one hand, the successful implementation of autonomous vehicles holds great allure for society as a whole. Autonomous vehicles would allow drivers to become passengers, freeing up time otherwise spent doing nothing else but driving.
This would allow drivers to be more productive, recreational, or even sleep during the daily commute to work. Moreover, it would enable for large fleets of self-driving cars that enable efficient ride sharing. Looking ahead, a society with fully automated cars would allow for faster transports that never get gridlocked, as well as fewer highway lanes.
On the flip side, one first needs to clear some significant hurdles. The perhaps most obvious obstacle to the successful self-driving cars is the technology itself. Whilst the technology has evolved a lot during the last few years alone, it still has a long way to go.
Currently, a self-driving vehicle cannot improvise or interpret new situations. This is why extensive testing is still needed. Furthermore, the advent of artificial intelligence (AI) and machine learning (ML) allows for cars to learn from the testing how to share information between themselves and to handle a myriad of different events.
This is sometimes referred to as a “Neural Net”. However, huge strides are already being made in regards to the technology required. The main barrier to the widespread use of self-driving cars might, in fact, be people.
Hacking risks worries
Two-thirds of Americans are uncomfortable about the idea of riding in self-driving cars, according to a Reuters/Ipsos opinion poll.
In another survey, French insurer AXA found that more than 90% of today’s roadway deaths and injuries are due to human error. Therefore, self-driving cars would naturally be a good way to eliminate the risks. However, one challenge that stands in the successful widespread rollout, rather than physical safety, is the computer operations security aspect.
As cars become increasingly connected and digitized, they become more vulnerable to hacking attacks.
Matthew Channon, an insurance expert on driverless cars from Exeter University, has been sounding the alarm on the real risks of connected autonomous vehicles. This comes after several high-profile hacking cases, where hackers have been able to disable breaks or even remotely control cars.
These risks increase exponentially as cars approach full autonomous status. However, as the issue is becoming increasingly prominent, there is a large amount of work going into securing the integrity of autonomous cars’ systems, making it dramatically harder, if not impossible, to remotely compromise them.
While the Tempe incident understandably has many worried about the future of autonomous driving, the Association of British Insurers (ABI) is defending self-driving cars. A spokesperson for the ABI pointed out that whilst autonomous driving is still in the early stages of testing, most crashes are still due to human error.
Though, even the Uber vehicle in Tempe had a human driver present, the driver still failed to prevent the crash. Clyde & Co law firm partner, Nigel Brook, pointed out that a major advantage for autonomous vehicles is that they continuously record, collect, and share masses of data even when involved in accidents, allowing them to avoid future incidents.
Autonomous cars present some major hurdles to clear – however the future potential is too great to stop exploring the technology.