We not in Kansas any more.

So previously I mentioned why I thought self-driving or computer driven cars are a truly great idea. I restrained myself to only talking about the benefits within the current framework of driving. In this post I am going to show how, by adopting computer driven cars, we can go far further than this. I will do this by going through some current paradigms about driving to see how a switch to a computer driven car affects them.


"There are too many cars for the roads."
This certainly seems true as any rush hour commuter can testify. Crawling for an hour in bumper to bumper roads does not exactly shout out "more cars please". However the theoretical capacity of roads is much greater than what we use. Even so called "bumper to bumper" traffic normally has a few metres between cars. The amazing reaction speed of computers means they can drive far closer than the current standards. This higher car packing especially at higher speeds greatly increases the rate of car flow roads can handle. Switch to computer driven cars and we would easily be within the operating capacity of our roads in most cases. 


"You need a passenger."
Cars massively extend your personal freedom by increasing your possible daily range. However, as I mentioned in the last post driving is a serious responsibility. Many times you would like to drive somewhere but when there you also want to do something that would reduce your ability to drive (be awake a long time, drink, etc).

We already know from Google's tests of their computer driven car that their car does not need any driver input so why require anyone in the car at all. Just imagine for a second how utterly amazingly useful your car could be if you could order it to go by itself somewhere. For example, say you missed a postage item it would be nice to send your car out whilst you are at work to collect it, would it not? Need to collect your in-laws from a distant airport in the early hours of the morning? Again no problem, send the car. Got lost in a bad area of town? Relax, just call the car to come get you at your phone's location (your own personal free uber). Work somewhere where parking is a nightmare, fine, just get the car to drop you off and go find a parking space then come back at the end of the day to collect you. I could go on and on but you get the point. It would be great.

The only drawback I can think of is the possible use of the car as a weapon or weapon delivery device. However the nature of computer driven cars would mean lots of logging which would mean your chance of getting caught would be very high. Cracking any software to prevent tracking will also be possible but if the criminals are that sophisticated and committed there are a million easier ways to attack. Hence, whilst computer driven car crime might be possible, I doubt it is really a practical investment in terms of effort. Basically, it would be easier and safer from the evil-doer's point of view, to either force an innocent person or manipulate a gullible individual into doing the driving as they do today.  


"You need a destination."
Seems obvious but it is wrong, actually you only need a desire. If I go to a scientific conference I skip the inevitably awful arranged site seeing tour and go for a little potter around the area on my own. In the past this required some pre-planning and possibly some purchasing of maps. Now I whip out the phone and search for nearby places of interest and away I go. If you are in your computer driven car why not do exactly the same "Car, take me to the best rated independent cinema in this city" or "Car, take me to the nearest place of interest that I might like". Naturally, sometimes you do need a specific destination; take me home is pretty clear to all but the very richest. For many other times, however, I suspect your destination could be far more flexible than you might think. We all like to think our interests and tastes are unique and eclectic, but as any big data advocate can tell you we are all very predictable really.


"Driving is not a green mode of transport."
This is true to some extent. Moving around will always require energy. However, computer driven cars can be safely driven close enough to be considered aerodynamically equivalent to a train. This coupled with their smooth driving style means the fuel consumption can be dramatically lowered. Their driving behaviour is also tuneable to take maximum avail of the advantages of alternative power sources such as electric. 

Electric cars clearly are the future but they currently have a serious problem when it comes to cheap battery storage capacity. One possible solution is to embrace some sort of swappable battery scheme. However, such a system is a pain as forgetting to do a "battery run" is highly likely. However, what if you removed that possibility and instead, every couple of trades, after your car dropped you to your destination it asks can it take ten minutes to pop over to the garage to go a battery switch. Provided the swaps were not too frequent and at least semi-automated (as in a garage attendant might be needed but not a driver) I doubt many people would mind. Of course this also requires a nation-wide system but then so did petrol or diesel cars when they were introduced.

Long journeys with electrics car would be easy as the car could automatically add several battery swap points when it calculates the route. Breaking an eight hour marathon drive into four two hour drives with three five minute breaks in-between. If that sounds annoying remember you might get half the CO2 emissions and you can sit back and watch a film with a beer or even go to sleep the whole way.


"You need to stop at traffic intersections."
Why do you need to slow down? Well there are two main reasons. One is that for traffic intersections with corners taking it faster would be dangerous or just plain unpleasant due to the acceleration. Computer driven or not, physics is physics and so you will have to slow down. However, it is safe to say that the vast majority of bends are taken at far lower speeds for another reason. Unless you drive on almost empty roads you need enough time to calculate the path other drivers are taking through the junction and thus safely avoid them. Computer driven cars can do this faster than you so they have less need to slow down. As it stands they are however coded to still slow down for the benefit of the nearby human drivers.

If however there were computer driving only lanes we could go further and do far better. The computer driven cars could broadcast their short term destination as they approach the junction for example "car 51587311 doing 32.5mph 800m from junction 212 taking 3rd exit". All other nearby computer driven cars would then receive all the similar transmissions from other computer driven cars. Each could then plan ahead to filter into the future gaps left between each other and rebroadcast their updated trajectory. This is effectively the complex version of dealing with many lane to a few lane mergers (as you often get at some US toll booth plazas) where as you emerge you look left and right and adjust your speed to slot into the gaps in the cars in the merging lanes. Actually, when you play top down driving games, like early versions of grand theft auto, humans can also do this at normal junctions. However in real life a human driver cannot normally see or safely predict the trajectory of other cars before entering the junction. Even if they could listen to the computer driven cars' broadcasts it is very unlikely they could visualize the state of the upcoming junctions beyond the trivial case of an almost empty junction. Hence, no matter how good a driver you are, very rarely is it safe to take a junction without greatly slowing down to assess the nearby cars' trajectories. With short term destination broadcasting and the predictability of computer driving cars a computer driven car can do this quite safely.

As the computer driven cars would be monitoring their neighbours, as they traversed the junction, it is likely they could tolerate and avoid some human drivers (or malfunctioning computer driving cars) in their lanes and junctions. This is especially true if they broadcast warnings to other computer driven cars (and possibly the police) about nearby "humans/malfunctioning-robots". Of course, the more you tweaked the junction driving algorithms to deal with trespasser  the more the efficiency of the junction would drop. I suspect you would start at a very robust safe system akin to normal human style junctions and increase the efficiency as confidence grew until the computer driving cars started reporting near misses (proper incident reporting is an massively under acknowledged advantage of computer driven cars). Traversing one of these higher efficiency junctions would be pretty amazing. You might need to close your eyes the first few times until you got used to it but given both journey times and fuel consumption would be greatly reduced I suspect you would learn to accept this quickly.


"Jams are natural they can't be prevented"
In some cases this is true for example unforeseeable accidents. Truly unforeseeable events such as sudden mechanic failure are very rare but do happen. You can argue that computer driving cars with their more constant systematic interrogation of car telemetry would be more likely to pick up on subtle sensor signals and provide warnings in advance. Maybe they might pull over and point blank refuse to drive if a dangerous state arose. However, things will go wrong as no system is perfect. Crashes and Jams will happen. The real question is now to mitigate their effects. The answer is rapid inter car and system wide communication of the jam. The best outcome is for all observers to quickly notify (incident reporting again) emergency services, the routing system (eg google maps) and nearby cars. That way the emergency services arrive as soon as possible, cars heading this way sufficiently far back enough can be automatically rerouted, and nearby cars can slow down and file into any free lanes.  Of course this may not avoid a jam but it will limit the jam as much as possible.

Most jams are not unforeseeable, they arise quite predictably. Even the apparently mysterious car jams that you drive into, crawl through slowly and emerge from the other side into open road without any hint of what would have caused the jam are predictable. Often people will mistakenly assume that these mysterious car jams were caused by some accident which has been cleared and the jam is just a remnant. It is actually more likely they have just traversed what is known as a Jamiton. Viewed from a distance, traffic jams often look like waves on the roads whose shape and size can be relatively constant over hours as it travels along roads at vastly different speeds to the cars going through the jams. This wave structure is called a Jamiton and it is what is known as an emergent phenomena. 

Like most emergent phenomena (other examples include Cities emerging from human settlements, colonies from ants or crashes in stocks exchanges) Jamitons do not appear below a crucial car density but go above and they emerge time and time again. The likelihood of Jamitons can be adjusted by subtle changes in the rules that govern the basic elements (cars in this case). Armed with this knowledge, traffic controllers often tweak local rules to attempt to suppress Jamitons. The most common way is by temporarily reducing the speed limit leading up to the Jamiton. However as many a sad driver knows this rarely works. The reason being drivers know most drivers will ignore the temporary reduction in speed limit. It is ignored for a variety of reasons but chief among them is a herd mentality that it is okay as everyone is ignoring it. This greatly reduces the traffic controller's ability to counteract Jamitons. Computer driven cars will obey the local law or even more subtle behavioural requests (keep an extra 3 m space to the cars in front for the next 4 km). With such obedience Jamitons can be effectively suppressed leading to more effective road use and less jams. 


Summary
This all might seem outside the box thinking but to be honest it is not that far outside. Every one of the ideas mentioned above has well established research proving its validity. None are cutting edge science. Implementing them is of course a serious engineering challenge but not a herculean one. Really the only thing stopping their widespread implementation is the will to do so. Partly, this is because before Google's computer driven car research no one was really thinking about it and partly this is because we have hopelessly fallen for the driving illusion of self-control and freedom. It is this illusion that leads to the false sense of ability I highlighted in my last post on automatic driving. Driving off into the sunset is an age old romantic way to end a film.

I for one would be happy to sit back with my feet up and a mojito in hand while the car drives itself and me into the sunset.