Skip to main content

2013 trend #3 - self-driving cars

In my first posts about 2013’s trends I covered wearable computing and 4G in the UK. In this third instalment I’ll look at the self-driving car, a long predicted technology that is finally ready for the limelight.


What are they: pretty simple to describe! A self-driving car doesn’t need a human driver to get from A to B. The driver tells it where to go and the car does the rest, navigating, dodging other road users, swearing at bad driving and doing burn outs at traffic lights.

My prediction: cars that self drive under certain circumstances will go on sale in 2013, starting with Mercedes’ flagship S Class and Audi’s A8 executive saloon, both of which are already announced. These first generation self-drivers can operate autonomously on the motorway, but not in urban areas or back roads. Second generation self-drivers, able to operate anywhere, are probably for 2018 or 2019’s models.

Why I think this: Plenty of cars on sale in the UK today have elements of self-driving technology in them. These “generation zero” elements are technologies like self-parking, collision avoidance and adaptive cruise control.
The chart shows the proportion of cars launched in the last 3 years on which these technologies can be specified as an option. What Mercedes and Audi will offer next year is integration that make these things work together as an autopilot under controlled conditions.

Motorways represent a simple driving environment. They are relatively straight, the chances of unexpected events like a child running into the street are extremely low. Furthermore, traffic speeds are relatively sustained and predictable. With lane following sensors, adaptive cruise control (which maintains a distance between the car and that in front, rather than a set speed) and collision avoidance systems (automatically stops the car when objects suddenly appear in front of it) the in car computer can take control of driving in traffic or on open roads.

All of this is made possible by other fundamental changes to car architectures that have happened somewhat under the radar of the average consumer. Electronic power steering and throttle control have replaced direct mechanical control, principally because they are a means of reducing fuel consumption.

In most new cars, turning the wheel or pushing the throttle is actually an instruction to a computer to do something, rather than a direct link to the wheels and engine.

Full self driving cars are more challenging. As Google have ably demonstrated in Las Vegas it is perfectly feasible to create a package of processing, sensors and hardware that enable a car to drive on any road, recognising signs and navigating the street.

Rumour has it that Google’s technology package costs $200,000. I’ll get to the money in a minute, but first, the tech’. I’ll make a list, because I’m in that sort of mood:

  • GPS – navigation systems are a common fixture in today’s cars. All but one model launched in the UK this year and last could have integrated navigation as an option. This is a fundamental technology for the self-driver as it enables long range journey planning (i.e. the driver to tell it to go to the shops and the car to know where that is). It also enables the self-driving software to “know” what the road looks like. Warnings of traffic lights, street furniture, speed limits, traffic and so on are all features of the latest navigation systems. Having foreknowledge of these things will enable the computer to be alert to danger zones.
  • Street view – Google’s street view and equivalents are a useful part of the car’s situational awareness suite as they offer another perspective on the roadside environment. What would be even more useful is the use of a live street view from the cameras of cars navigating the same route that give warning of unexpected roadside situations – road works, for example.
  • Data connectivity – many mid-to-high end cars are available with cellular data connectivity built in. If there were a single web service that enabled interchange of the data they collect on traffic conditions, street views, the position of other cars and so on, then the practicality of self-driving cars would increase considerably. Frankly even if a major manufacturer like VW did this for its group then feasibility would increase.
  • Machine vision – can read road signs and use the results to determine likely hazards, traffic laws and so on. Can also detect red lights, which is useful. Some cars – Ford’s Focus, for example – already feature image recognition technology that flash up speed limit warnings on the dash when they see a sign beside the road, but the feature is not yet integrated with safety systems
  • Lidar – or laser radar – uses a rapidly rotating laser to scan for objects and thereby builds a 3D picture of the environment many times a second. This is an intermediate range warning system that enables the car to understand where objects are in a 20-50m radius.
  • Collision avoidance/ adaptive cruise control – these systems are based on either radar or ultrasonic systems and are a last ditch defence against objects that unexpectedly appear. The classic emergency stop by the car in front challenge.

I understand that Google’s $200k self-driving package incorporates most of the above technology. This seems like a crazy amount of money, but unlike really cutting edge tech such as fuel cells, the Google package is basically all standard computing and sensing technology. The Lidar is the only low volume component.

If you apply Moore’s Law to that cost, you end up with the complete package costing $12,500 by 2018. That’s still quite a lot (and assumes quite a lot!). The other thing that’s worth considering is that once the algorithms that enable a car to self-drive are understood, they also improve at a considerable rate. So in reality the computing gets cheaper while the problem gets easier.

The minimum I’d expect is that the cost of providing the complete package of tech would fall below $20k by 2018. It could be a lot lower and doubtless will be if there’s widespread adoption amongst automobile manufacturers.

Implications: Self-driving is hugely beneficial for society. Computers are simply much better drivers than people. Although the very best human drivers might be faster than a computer, they still get tired and are randomly fallible. If the majority of cars were computer controlled then accident fatality rates would fall – possibly quite substantially.

Furthermore, human drivers are pretty poor when it comes to avoiding jams and driving fuel efficiently. Computers are much better. So emissions would fall, journeys would – on average – be faster.

Productivity and general wellbeing would certainly improve as well. Commuting time is not particularly productive time and it’s often (but not always) relatively stressful.

But there is a crucial hurdle to get over. Very few people would feel comfortable with letting an invisible driver control their vehicle, at least initially. In most parts of most of the world self-driving cars are illegal as the driver has to be in control of the vehicle. Whether legal systems could accommodate the challenges of computer glitches causing accidents (as opposed to far more common human glitches) is another question entirely.

Google’s experiments in Las Vegas graphically demonstrate the feasibility and benefits of self-driving cars. If human reticence to believe in the power of the machines that they’ve created can be managed, then the rest of the world can soon enjoy the same upsides.

Comments

Popular posts from this blog

Impacts of a handset leasing model on mobile telcos

Following yesterday's post, here's some related thinking on the impacts on operators of handset leasing. Handset sales represent around 25% of operator revenues in a typical European market, but generate only around 5% of margin. It may therefore be the case that the scenario described would lead operators to a more profitable structural model than exists today. Oil companies are consistently and acceptably profitable, despite being (literally in some cases) the ‘dumb pipe’ that operators are so desperate to avoid becoming. One of the reasons for the oil majors sustained profitability is clear focus on their role in the value chain – to supply the fuel that enables transportation, relying primarily on location, then brand and finally product innovation to compete. BP or Shell do not need to subsidise the purchase of a car in order to drive consumption of fuel because consumers are ‘hooked’ on it (it gets them from place to place) and there are many credible car manufacturers an...

Differences between Industrial and Digital businesses

Since I'm stuck on a Eurostar crawling through western France I thought I'd use the downtime to share this table I've made on the differences between Industrial and Digital companies across the main business functions. A strange insight into how my mind works... but hopeful a useful summary!

Value drivers for telecoms retail

I've been doing a really large number of driver trees recently - we've taken to using them on every project to get really into the guts of value creation for businesses and thus decide where to focus initiative development (How To Win, if you're keeping score). Anyhow, I had to pause for thought recently to work out how to represent the subscription aspect of telecoms retail for a client. Since it took me a minute, I thought I'd share... its lack of elegance suggests that its not quite right, although it was enough to demonstrate that there was a certain lack of coverage in the initiatives that my client was pursuing and thus spark a debate. Enjoy.