Building confidence in driverless technology

Driverless cars have been big in the news for some time. But, recent news stories this month have brought renewed focus on the issues involved.

First there was the Google driverless car involved in a "fender bender" with a bus. Then a few days later it was announced that Britain was grappling with the issues of driverless cars.

MIRA cooperative vehicle

The third news announcement was the National Transportation Systems Center in the USA, concluding that increased levels of automation are acceptable provided the cars have controls such as steering wheels and brake pedals.

So much for Mercedes-Benz driverless lounge concept which, although it had the option of a driver cockpit, also allowed the driver to swivel the chair round to face into the centre of the vehicle or "room".

The wariness with which the authorities are approaching the issue of driverless cars may have something to do with our understanding of computing. A computer, as we know, can only give binary answers – 'yes' or 'no'. It has no ‘maybe', ‘perhaps’, or ‘I am weighing up the situation’.

Of course, a computer can process so many binary questions in an instant that it can be given complex problem-solving capabilities by using a myriad of ‘what if' binary choices.

But we find it hard to anticipate a computer that can analyse the situation and decide: 'that driver is driving erratically, therefore I need to be cautious'. Or the even more elusive gut feeling ‘that driver is about to drive out in front of me’.

Even on the more basic automation questions, I still have concerns about the fundamentals.

For example, will a driverless car drive me off the road? I have used lane-keeping technology and I don’t fully trust it. I worry about road markings.

It's all very well combining lane-keeping technology, with adaptive cruise control and radar, but what happens when your lane-keeping camera comes across bad temporary road markings? Or no markings at all?

For all these reasons, I think the slowly, slowly approach is best.

Judging by insurance premiums, the idea of cars with autonomous braking systems – where the radar will apply the brakes to avoid, or mitigate, a collision – is accepted as a real safety benefit. The insurance companies have confidence in the systems.

Having driven cars with such a system (including my own) so do I.

The next logical step is Traffic Jam Assist, which allows a car to drive automatically in slow-moving queues automatically steering, braking and accelerating to maintain its position.

Tesla went one step further with their Autopilot software which was made available last autumn.

The ‘Autosteer’ module is clearly marked “Beta”. Tesla had to remind people that the software was still in test mode and that “Tesla requires drivers to remain engaged and aware when ‘Autosteer’ is enabled. Drivers must keep their hands on the steering wheel”.

Despite these cautions, the internet started filling up with stories, photos and videos of near misses as drivers put the new software to the test. (There were almost as many photos, videos and accounts of it working well, but it is always the negatives that gain the highest profile.)

The ‘Tesla Autopilot tried to kill me!’ stories and videos are just what the self-driving car movement does not need.

All of which suggests a more cautionary approach that allows the public, the authorities and the legislators to get comfortable with increasing automation.

Driverless car

I’m comfortable with autonomous braking. I am sure I would be comfortable with Traffic Jam Assist and in the future I expect I may have a car with that feature. I can see ‘Motorway Assist’ being something I would be comfortable with in the not to distant future. I’m not so sure about ‘B-road’ assist though!

Next year Nissan will start production of a Piloted Drive Qashqai at their Sunderland plant in the UK. Nissan say their ‘Piloted Drive 1.0’ “allows cars to drive autonomously and safely in a single lane in heavy traffic conditions on highways”. The company continues to say that, over the next four years, Nissan will launch vehicles with increased autonomous capabilities.

Meanwhile Ford says it is tripling the size of its autonomous vehicle development fleet, while Jaguar are testing technology on a 41-mile section of connected and autonomous vehicle roads around Coventry and Solihull.

What we need to see is these new capabilities being introduced and being accepted. Self-driving cars are on their way, but we need to build confidence in them. There is some way to go yet.
blog comments powered by Disqus
© 2007-19 Ken McEwen. All rights reserved. Contact me