left1I see a lot of posts about automated or self driving cars in my Linked In stream, probably because of all the QNX people I follow. One of the main areas the QNX operating system is used in is automotive systems so that’s not surprising. And Google is also frequently in the news for their self driving cars, so the topic is really starting to get a lot of attention.

I have no doubt that we’ll eventually have cars that can drive for themselves. We’ve seen them demonstrated and it’s a really cool idea but I still see a lot of problems with them.

The biggest problem I see is that they will need to deal with unpredictable humans who don’t follow the rules of the road. If we program self driving cars to ‘believe’ the turn signals on other cars, what will they do when they meet the stereotypical Sunday driver that forgets to turn off their turn signal?

What will they do when the car signals a lane change, but a rude driver won’t let them in?

I think it will be a long time before we allow computer driven cars without a human override. What I think makes sense, and what is already happening, is that we will allow our cars to incorporate safety overrides at first for things that it makes sense to do and that we can’t do as humans quick enough to do effectively or reliably. We’ll let them take over driving for us once there is no doubt that we’re in a situation we can’t deal with.

Call it ‘augmented driving’.

Two examples of that kind of thing are airbags, and anti lock brakes. Bracing for a collision is something we have always done and an airbag is just a fancy robotic collision bracing system that auto deploys (hopefully) only when needed. Pumping the brakes manually in slippery conditions is something we’ve always done. An anti lock braking system just does it better than we ever could and only when needed.

So I think the evolution from here to fully driving cars will be very long, and will take place in incremental steps like that. We will start by instrumenting everything that’s important so that the cars can sense when important things happen, and we will equip them with specialized information displays, annunciators, actuators and software to take over at the right time and override us.

The rear cameras that tell people when they are backing up and there is an obstacle are a great example of that, and so is the parallel parking mechanism being deployed in some cars. Collision avoidance radar will eventually be turned into a system that actively veers to avoid a collision for you if someone hasn’t already done that.

I’m personally trying to promote the idea that smart cars should  sense when some living thing has been left in a hot car during a heatwave and actually do something to prevent harm. That’s pretty low hanging fruit.

There is plenty more of that kind of assistive technology coming down the pipe very soon, but it’s a co-pilot we’re building in now, not a pilot.

It’s likely that for a while, there will be special cases where roads are built for the exclusive use of completely self driving cars and we may develop an infrastructure for commercial traffic that uses them. We will need ‘smart roads’ for the concept of self driving cars to be fully realized and it will take a long time for them to be built. It will likely start out being done in small islands of experimentation before it gets rolled out and adopted for all roads.

In the meantime, we will continue to augment the reality within human piloted vehicles with more and more information systems and technology. Big data will probably be used to collect information about how people in the real world actually drive and run it through some fancy AI – maybe IBM’s Watson – to come up with come kind of neural net ‘driving brain’ that could allow a machine to drive in much the same way that a typical human does, even on regular non-smart roads.

I just hope we don’t train them to leave their blinkers on.

Comments are closed.