Robocars: The Roadmap to getting there


In part two of my series on Robocars, let me introduce:

The Roadmap to Robocars

Here I outline a series of steps along the way to the full robocar world. We won't switch all at once, and many more limited technologies can be marketed before the day when most cars on the road are computer driven. Here are some ideas of what those steps could be -- or already are.


While there's few things I'd love more than step into a car and not worry about driving at all to work - it's not driving for pleasure at all, but a stress- and congestion-filled mess - the bigger issue is that right now, the US is very much based on the idea of cheap fuel. It's not just the efficient robocars that have to change the transport landscape, but also the infrastructure.

Changing the infrastructure and allowing people to use a bike or to just walk instead of infrastructure being hostile to these forms of transport would also do good - because it'd eliminate the need for a car for smaller distances. The 3-wheeled Aerorider you showed can negate concerns about weather and carrying things - it's just that the cost is much higher than that of a conventional bike, which couldn't work as smoothly in regular traffic due to its lack of speed.

edit: Should've read the "Objections" page. Since the roads are already very wide, just painting lines and making laws protecting cyclists might do the job in the meantime.

I suspect that we can get much of the way towards the robocar vision without actually admitting it. That is, as you described, robocars will be sold as regular cars with crash-resistance built in as a feature. Then, as people realize that they can tune out while "driving," they'll start paying less and less attention. At some point, people will admit that they are actually passengers and not drivers, but at first it will be, "I'm just letting the crash-avoidance system take over for a few seconds while I find my kid's sippy cup." Those few seconds turn into minutes as people learn what they can get away with.

I hope you will consider joining the Advanced Transit Association ( if you haven't already. Some of our members promote your ideas and are developing them. I haven't seen a link here to, which gives a lot of links to European companies and research. There are some cybercars/robocars in operation in Europe.

One of the overall problems is that the intermediate stage (ie mixing Robocars with human driven vehicles) is going to be much more difficult to acheive than the end stage of Robocars only(particulaly from a safety point of view).
I am a strong believer that robocar will happen. But despite the excellent arguments you make in your road map, I think that the ability of Robocars trying to avoid accidents would be unable to compete against the ability of humans to cause them.
Also in the early years of development, robocars will need to travel at a slower speed than humans and this would cause too much frustration.
I think that it will prove necessary to install a robocar network that uses roads free of general cars.
The concept could be first developed in a large test area (perhaps a gated community beside a major university) and then set up as a hub and spoke model in a major city (maybe a very congested European city running out of tranport options).
The hub could be the CBD and the spokes could run out to residential areas and car parks. This I think would be a more technically acheivable option than putting humans and computers together on our roads.
Humans can cause computers to crash just with a few key strokes, imagine what they could do to them with a 2 ton SUV.


The problem is that developing special road networks is expensive, and banning human drivers on them is really expensive. Especially when there is a chicken and egg problem. Few robocars exist, so what justifies their dedicated lanes? Few dedicated lanes so who buys an expensive (at first) robocar that can only do what it does best in a limited area?

There are some possible intermediates. For example robocars on highways (with human drivers) but not on city streets. Or possibly vice versa, because possibly slow-speed complex driving is easier to solve than high speed simple driving because reaction times are better. One can't say.

But I think the challenge must go out the engineers to be much safer than humans on today's roads. That vehicle then is an easy choice to buy as it can go everywhere that it's legal to go in one, ie. your whole city or whole state highway network or whole interstate system etc.

Your approach seems easier, but the question is, which happens faster -- the political and financial changes necessary to build a large network of dedicated robocar-only roads, or the robocar engineers working so as to not need that?

Cars that drive themselves could be made to communicate which would allow close drafting on the highway greatly reducing highway fuel use.

The school of fish test seems MUCH too high a bar to pass, and it's completely unreasonable to expect robot cars to be able to pass it, when not even the best human drivers have the sligthest chance of success. In heavy traffic, if one participant is actively trying to bumpt into other participants, he will manage it, without much trouble at all.

This is especially true if the test isn't limited to low speeds, if there's head-on traffic, or if the possible separation is low, for example if the streets aren't very wide. There's not much anyone can -do- about for example a overtaking vehicle on a narrow straight road, that suddenly decides to turn full into your side while alongside you.

Being demonstrably safer than the alternative, should be enough. It makes no logical sense to allow human drivers, with their proven track-record, but disallowe robotic cars that demonstrably are safer.

I do write that head-on is too difficult and we might move to having mostly one way streets for a number of reasons.

As for the sudden side turn by a car overtaking you, I have not proven you can always handle that but you can handle it a lot of the time because your reaction time is near instant and you can brake and turn as well as the other car.

Also proposed has been the idea of deliberately hitting other cars, at low speeds, where a group of robocars, noticing a car doing something that could lead to an unavoidable bad collision, work together to make it a soft one, by surrounding the the bad car and constraining its movements. Ie. if a car starts turning towards you and there is no path out, you turn towards it and sideswipe it at low relative velocity to stop it from turning more sharply.

Add new comment