Wednesday, December 19, 2012
Thoughts on the driverless revolution:
Andrew D Atkin:
This article is essentially an extension from my previous piece on full-automation transport.
Google's driverless car has been proven, and Google has already got the state's of California and Nevada on board with them. They are now working with officials in these districts to develop laws for the implementation of fully-autonomous vehicles.
Google says that it now needs to move forward as fast as possible with this technology, which I believe can be decoded as: "All our competitors know as well as we do that driverless cars are the soon-to-be future, because we've shown that it can be done and that we will continue to develop this technology. And most importantly, the key political barriers have been removed". [note: This paragraph is not a literal quote].
In other words the race is on. No-one informed on this issue is second-guessing whether driverless cars are the future now. Even politicians, world over and in New Zealand, are starting to talk seriously about it. Good!
What driverless cars mean:
Most people think in terms of a new system under an old roof when thinking about driverless cars. It's wrong-thinking, because driverless cars will redefine transport as we know it and even much of the operational structure of our cities.
Firstly, most of us will not personally own a driverless car. Driverless cars will give us the rise of the taxi, though the demise of the taxi-driver. They will give us network-based transport whereby we simply order-up the car we want, for the job we want, and then ditch it (or should I say it ditches itself) after you've used it. Network-based transport will be the 'new roof'.
Driverless cars will be directly matched to demand, which means that most cars will be designed for single-occupancy usage only. This shift, more than anything, will revolutionise transport efficiency.
So how should the single-occupancy vehicle look?
The most substantial thing you can do to make cars more comfortable, is to design them to be self-tilting so as to get rid of side-forces while traveling. Doing that in a 4-wheeled vehicle is problematic, but in a 2-wheeled vehicle it is of course easy and necessary.
This is why I predict that the most common type of autonomous taxi will be a somewhat spacious, enclosed motorcycle. It will be stabilised with retractable wheels (for static stability) and/or gyroscopes*.
The following image gives us the Mono-tracer. It's an enclosed motorcycle similar to what I suggest.
Also it's important to note that with driverless cars, the free time that's liberated makes comfort only more important. If people are to use their cars as an office, dinner table, bed, etc, then getting rid of side-forces will be even more important. This is especially the case for a hilly country like New Zealand which is almost nothing but corners.
And finally, an enclosed 2-wheeled vehicle is extremely economical and energy-efficient to run. With aggressive streamlining, you can get a hybrid vehicle of this type down to less than 10% of the fuel consumption of a standard 4-wheeled vehicle today.
Google has demonstrated that it can fully automate its cars with no infrastructural investment in the existing road network to accommodate them. That is quite a testimony to how far this technology has come, and how close its implementation is. However, because we now know that driverless cars are the future, it might be to everyone's advantage to look at the possibility of directly conditioning the roads to assist electronic control - for the sake of reducing net costs, and possibly improving performance.
We could look at impregnating passive-RF chips in the ground, so auto-cars can get instant and totally exact feedback on their positions (with no advanced information processing), and we can maybe install embedded wires that contain a radio signal for easier positioning of the cars, etc. There is also the possibility of installing road-based sensors that communicate to auto-cars, allowing them to see around corners. This would radically empower the defensive-driving capability of auto-cars, and far beyond what a human could achieve.
Again, because we know where we're going with this technology, it makes sense to study how best to accommodate the driverless revolution via the public-infrastructural response.
Although there are many important questions around how our cities could (or should) develop in response to driverless car technology, if we can loosely predict that the enclosed motorcycle is the future then the government could consider the following:
The idea is to condition existing roads so as to provide discrete super-elevation for autonomous 2-wheeled vehicles. The hump in the road (red part of the included diagram) would only be about 1 to 2 feet wide.
On the scale of things it would be a trivial cost and an easy thing to do. It would make autonomous vehicles safer, more efficient and significantly reduce tyre wear. Also it gives us the opportunity to smooth-out the roads for most traffic, further improving comfort.
The motorcycle would have to follow a relatively strict travel path on the road, but that's easy enough with electronic control. However, slight rear-wheel steering might be necessary if super-elevation is used in tight corners.
-Further to improve smoothness and comfort, the bikes should be series-hybrids. This means make them electric, but include a diesel-generator for range. Care should be taken to isolate vibrations coming from the generator, and to this end an opposed-piston engine would be very effective as it's almost perfectly balanced (vibrationless).
Local governments in New Zealand are infatuated with the belief that their cities should be high-density in form, and they in turn believe their cities should be forced to evolve in that way. Ignoring the fact that they are desperately wrong, the fact remains that full-automation transport technology is a radical game-changer on its own, no matter what your current planning philosophy might be. Local and central governments will need to rethink their ideas in relation to a full-automation transport world.
For example, if people can commute on non-congested roads at 10% of the total cost of today's transport (not unrealistic), and do business in their car via the Cloud, how then would this affect the rational structure of a modern city form? Do we really need to suffocate land supply so as to force people to live closer together, when it's just as or even more efficient to let them spread out? And when such beautiful locations to live in are so easily accessible, in terms of both travel time/pleasure and cost, does it then really make sense to actively deprive people of this option? And how will your economy fare by taking away people's idealised lifestyles, while other economies do not? Will people even hang around?
Further still, if people can work from home yet easily access their work place for when they specifically need to, then how many new commercial buildings will we need to build? Maybe New Zealand would be better off just demolishing its dodgy earthquake-risk buildings and rolling-out the fibre-optics instead?
There is strong rumour that Google is already working to implement their autonomous technology to provide a delivery service. It's obvious enough that a delivery service would be an ideal first-step application for driverless cars. A car that delivers goods may only need to be about, say, a 30th of the size of a normal car, so it won't have the strict safety issues associated with bigger vehicles that transport people. And the operational costs of tiny autonomous cars will of course be trivial.
Micro-cars will give us the "physical internet". The impact that micro-cars alone will have will be fascinating.
I think one of the most distinct effects that we will see from micro-cars is home-cooking being reduced to hobby status, because it will be soon be too easy (and cheap) to have a good meal delivered to you. With micro-cars the rationale exists to develop massive kitchen complexes that mass-produce varied meals, because the system allows for a single production-point to reach-out to a large consumer-base efficiently (and quickly). Also these kitchens can be efficiently supplied with fresh local food. Watch out for food production monopolies/cartels developing in the future.
The micro-cars also make it convenient for people to hire more than buy, for infrequently used items. And they can allow retail to come to you rather than the other way around. I predict that retail as we know it will be largely converted to a showcase industry - most things will be bought and delivered online.
Autonomous transport allows for efficient outsourcing in production. So efficient, that it can give us a "go-anywhere" production-line, allowing the production process to be efficiently split amongst various factories. The advantage is better utilisation of capital and skills for smaller production runs. Conversely, full automation allows machinery to be transported efficiently to a site, also allowing for better utilisation of capital. Hiring an expensive machine is easier if it's used twice as often.
The physical internet makes bypassing the middleman easy enough. Manufacturers will no doubt prefer to sell their products directly, as traditional retail is seen more as a parasitic cost.
Both the public and private sectors need to start thinking about driverless car technology, and how it might affect their operations. Driverless technology will penetrate society in ways far beyond what I have currently speculated over. It's not here yet, but it's close, inevitable, and the impact will be nothing less than profound.
The sooner we can know what we're dealing with, the better we can manage (and exploit) its future impact.
*The feasibility of gyroscopes is questionable to me, simply because of the bearing wear. Small gyroscopes must spin at an incredible speed for an application like this, and I don't know if they will prove to be efficient or practical for general vehicle use.
New Zealand tourism, in particular, could benefit enormously from driverless cars. This is because New Zealand is like the entire world under one small roof. It goes from snowy mountains in the South to sub-tropical in the North, within a total length of 1,600 km. Obviously if people can take a driverless (and self-tilting) car to where ever they want to go, without any hassle, fuss or significant discomfort, then this will be worth a lot to any traveler, especially a traveler who does not know their way around.
I have expressed before that I believe there is huge potential for New Zealand tourism exploiting driverless technology. Maybe it's time to focus on what this will mean. Realistically, the value of New Zealand's attractions are as great as they are accessible. Driverless technology can and will hugely improve practical accessibility.
Auckland and other areas of New Zealand are hell-bent on investing in rail systems, which they believe to be the future for transport. The modern polycentric city does not work too well with collective transport, which is almost always only effective for supporting CBD's of which, generally, represent about 10% of metropolitan transport demand.
Without going into details on this debate, I will point out that one of the first "casualties" of driverless technology will be the shuttle-bus, because this is where driverless technology for passenger transport will make the most sense to begin with (maximum bang for the buck).
So how will buses and trains compete with automated shuttle-buses that can platoon to form road-trains, as required, and can operate with more flexibility, frequency, are faster, and have a much reduced operating cost? Auckland is currently looking to invest billions into what is about to become a glorious white elephant rail system. See here for my article on automated shuttle-buses.
If Google plays it's game right (and I think they probably will) they will work with their driverless technology to become the one-stop hub for most material online transactions.
A driverless delivery service is about reducing costs and increasing convenience, and a logical way to reinforce this value is to provide standardised online formatting for consumers and suppliers who are using the automated transportation network.
This means creating a standardised web format (similar to blogger, which is what you're reading right now) so that anyone selling a product and/or service can easily put their products online with Google. Having a standardised format means that consumers don't have to "learn" a new website every time they want to make an online transaction with a new supplier, which of course makes online purchasing more attractive.
Another thing Google should provide is Google credit. Basically, the model I'm thinking of is to have the consumer install credit onto a 'Google account' in the same way that we install credit onto our cellphones. From here, within the standardised web format, people can make a one-click purchase instantly to any third party integrated with Google's system. Total streamlining - no time-wasting duplication.
Google has the golden opportunity to do this, because it's in the lead with producing an automated delivery service. They can compound the value of their services and accelerate the demand and evolution of them, by simultaneously developing this kind of online support infrastructure. I hope for everyone's sake they move in this direction as it only makes sense, and I would be surprised if they do not. The online world needs a fully streamlined hub like this, for efficiency.
Google has suggested that driverless cars will one-day cut the road toll by 50% or more. I have to say this assertion is beyond conservative, and looks more like playing it safe (by not over-promising) than realism. I would say they're probably claiming 50% for in case the worst happens in these sensitive early stages.
The truth is in a driverless world accidents will quickly become freak incidents. They will happen from things like landslides and tornadoes, and maybe the odd kid running out on the road at the very last second, overwhelming the stopping power of rubber-on-road.
...But try more like a 98% reduction - not just 50%!
The key to making fail-safe systems is redundancy. Build your system so that if one component fails there will be another to take over its role, to avoid an accident. You can achieve redundancy through both hardware and software. For hardware, for example, you can install a powerful handbrake that can also co-function as a fail-safe for the main breaks, if they fail, etc.
As for software, you simply programme fail-safe adaptive programmes into the computer. Think of driving your car down the road and the accelerator and brake gets stuck. What do you do? Turn off your engine, change gear for engine braking, and maybe apply the handbrake. This example scenario would not be a back-up in terms of hardware, but a "software" response to failed hardware.
Google will no doubt use very reliable components in all their cars, and in time they will surely develop appropriate redundancy - making auto-cars far safer than what humans are or could ever be. This is another key advantage of having a network-based system - you can afford to invest heavily in safety features for each working car, because each car is replacing about 20 privately owned vehicles, trivialising the costs.
Also, a driverless world means less cars parked up alongside the road. The result is that pedestrians can see the cars coming much better, and the auto-cars can likewise see pedestrians better. The result can only be much safer for everyone - not just the people in the auto-cars.
Auto-Taxi: How it may play out.
Let's say it cost $50,000 to fully automate a taxi that's used 24/7. (And $50,000 is more than generous. It will be much less than that, in time). Looks costly, but even at that price it will pay for itself in about 3 months from labour savings. This is why we can expect the first mass-uptake of auto-cars to come from taxi operators - huge bangs for the bucks.
The consumer will not see a major reduction in taxi fares during the initial stages of the implementation of auto-taxi's. The profit will go to taxi operators, but that is what we want because it then provides the profit-signal that induces more (and rapid) investment into auto-taxi's.
However, in good time the (existing) market for taxi's will be saturated, and nearly all taxi's will be auto-taxi's. From here, established taxi companies will compete against each other, and they will also compete against start-up's for market share.
This is the point where the consumer will start to get the pay-off. Because auto-taxi's are much cheaper to run, their fare-price can drop hugely before they fail to make a tangible profit; and indeed, this price-lowering will happen progressively over time*.
So, the competition and initial market saturation will drive down prices to where you would expect them to be, relating to basic supply costs and profits. And as a result of lowered prices, auto-taxi's will begin to compete with personal transport and proportionally claim a greater market share over land transport in general.
Prices will further drop (and performance will increase) as auto-taxi's increase in number, which allows the total-system to better avoid redundant travel from the empty-returning/sending of vehicles.
At some point of the evolution, either sooner or later, we will see the introduction of ultra cheap single-seat vehicles. [Note: I would recommend to Google to look at doing this more sooner than later, because the cheaper and smaller the cars are, the more of them you can get on the road, and sooner, providing a rapid increase in service appeal. You can't lose with small cars. In a massive 7+ billion world that spreads from poverty to wealth, small cars will always be the ideal for a specific target-market, somewhere].
As cheap taxi's proliferate, we will see them come to completely dominate modern transport demand. Network-based transport will be the new model, and the new transport world.
-Some people have said that we won't see a driverless revolution until some time like 2040. Nonsense. Once driverless cars get a foothold in the taxi-world, which should be only a few years away, this progression will go off like a bomb. It's all about price signals and the price signals are too strong.
Note: Respect that the only thing we need for a 'taxi revolution' to begin is the legal go-ahead for cars to at least empty send/return. People can still drive them if we insist (while inheriting the advanced safety features and platooning capabilities of self-driving cars, included). So, you could begin with small 2-seat taxi's weighing only about 500kg or so (ultra efficient), and the central advantages would already be achieved. Legal authority to ride without driving can come later, if need be. Again - that is how close this movement is.
This, the TWIZY, is the sort of vehicle Google should begin with. Only use a doored 20_kw version, and employ a ~7_kw diesel-electric generator as well.
*er...so long as "McTaxi" doesn't somehow bribe your government to enforce regulations that choke-off competitive supply-responses. You have been warned! Don't let our property disaster happen in transport too. Make sure your government works for the citisen, and not some monopolistic creep show.
Seeing around corners:
In an urban environment, which is lit, you can mount a simple camera on an existing lamp-pole, whereby the camera serves the function of detecting oncoming traffic in terms of speed and position - collecting and sending only the most basic information. This is all you need to optimise the safety of an auto-car system, and it should be extremely cheap to install.
There's nothing to it. Just a basic camera, a solar panel + battery, a micro-chip with minimal visual software (to detect and calculate the speed of oncoming cars), a blue-tooth type message-send (to the auto-cars), and finally visible lines on the road to provide a reference for the camera-assembly to calculate the speed of passing objects. Refer to the following image:
Being able to have a system that can see around corners for cheap, means you can put these sensors anywhere you feel you might need them. It will help to make the system bullet-proof as far as safety and defensive driving goes.
In thinking about how we might form micro-cars, it would seem clear that we would exploit containerisation. The micro-cars will pick-up and drop-off standardised containers, because this is the most efficient way to utilise the service. It would be like the old milk bottle system.
Containerisation maximises the efficiency of the micro-cars, as the cars don't have to wait for a human to empty them of their pay-load. They just drop off the container, send a text-alert to the recipient (informing them that their container has arrived), and move on to the next errand.
Also, with relatively simple robotics many containers can be stacked together, rapidly, to allow for efficient long-distance freighting (small cars/items have high aerodynamic drag relative to their volume*). Of course there would be a range of different container sizes, along with different micro-cars, but they would still be standardised so that different sized containers can be tightly stacked together.
The following image is my best guess for how a practical micro-car might be formed, as follows:
It could basically be described as "A vacuum-cleaner sized, full-automation forklift".
The container would be fully externalised so it can rapidly and easily clip on and off. The drive-wheels would be positioned on the rear, along with the heavy batteries to balance the system, with its variant loads. The steering-wheels would be positioned central to the total vehicle for maximum turning capability (providing a very small wheel-base), which is important for practical manoeuvrability indoors. The suspension and steering system would be structurally the same as a motorcycles.
There are of course many different way you could form these vehicles, but what will surely be employed, I believe, is a system based on containerisation.
Wheels: Also, with this system you will notice that there will be a strong design focus on the wheels for the smaller micro-cars. It is essential to reduce the unsprung weight of the wheels in a very small car moving at speed, because it can otherwise be too easily destabilised should it hit a rock or pothole, etc. There may even be room for the development of a flexible wheel that distorts (buckles in) in response to hitting a large obstruction, which can probably be achieved with a post-tensioned rim and non-pneumatic tyres (I'll spare you the details).
*Note that for intra-city operation, the micro-cars can drive extremely close to any leading car (as close as a couple of centimetres, or so), which will largely remove their aerodynamic drag within cities, giving them excellent range.
In a car it's hard to concentrate on important work because of the subconscious distraction of movement. Other than having a self-tilting vehicle structure, and electronically controlled suspension, the best thing you can do is have the riders body completely supported while working. The following is what I think is the best system to achieve that end.
Rather than sitting up to type, why not lie back and have the keyboard directly attached to your chair? For those who cannot touch-type, visual presentation of the keyboard can be re-represented on the monitor, using cameras. Your hand would be rendered as a ghost-image in the screen (over the keyboard), allowing you to clearly see the symbols you are typing on.
I could imagine this system being appealing on other contexts as well, not just cars. But the trick to making it practical will be to cut right back on unnecessary visual information-processing, so as to get rid of excessive latency between hand movements and their visual representation on the screen. Also, the keyboards should still be formed with "clear" tactile feedback for ease of physical orientation.
This system should be perfectly workable, allow for fast typing, and be exceptionally comfortable to use. Another advantage of this system is that you can change the symbols represented on the screen at will, which can be good for when you're using unusual programmes; and this functionality effectively expands fast-access functions on an otherwise small keyboard ie. one click and your keyboard has basically reinvented itself.
-Voice recognition would be a useful tool in cars, as well.
-By the time these cars are available, your personal computer will be your cellphone. The car, as a terminal, will simply expand from your phone.
-Note also that the in-car armchair can be mounted on pneumatic vibration isolators, isolating for both horizontal and vertical vibration. This should help a great deal as well, in making the car more effective as a moving office.
Operating a mouse is a precise job, and the movements within a car could make it strenuous. My suggestion to make operating a mouse easier is to employ a "voice mouse". Basically, in conjunction with a large touch-pad, allow for the function of pressing a button so as to lay a grid over the screen of your computer, which highlights all the links that can be clicked-on by placing a simple number/letter code on them. For example, going by the below image, if you wanted to click on "My computer" you would activate the grid and then call out "A3". This would be the same as clicking on "My computer".
Voice-activation should work well in this context, because voice is easy to use in a private car, and there are only a small number of voice-codes necessary to run a practical grid, so it becomes easy for the computer to differentiate amongst verbal signals i.e sloppy-speech friendly.
An ideal commuter car?
In my comments (below) I asserted that 2 and 4-wheel vehicles, not 3, will be the way to go with driverless cars. This was based on my presumption that 3-wheelers will not self-tilt. Well, Toyota may be changing my mind. Their tricycle (refer to the following video) self-tilts in a simple and economical way. It will not be quite as efficient as a 2-wheeled vehicle (other things being equal), but the difference between being 10x or 7-8x more energy-efficient than a modern sedan is not really a mass-market concern. Comfort is going to be more important.
An active-tilt 3-wheeler has the advantage of providing independent suspension, electronically controlled suspension, and better isolation of the rider from the vertical movement of the wheels. The 3-wheeler also provides rigid resistance to small movements from side-wind gusts. In short, it should be notably more comfortable than a 2-wheeler, while still providing excellent efficiency and tilting.
Toyota's format looks to me like an excellent contended for the mass-market commuter car of the driverless future.
Standardised core cars:
I would hope that with the driverless revolution we develop a core vehicle fleet, as cars become more exclusively a social utility than a fashion statement of sorts. The advantage of having highly standardised cars is of course that they're more efficient to manufacture, and more importantly it facilitates efficient mass-production style maintenance (including cleaning).
Cars can drive themselves onto the first stage of a production line, be explicitly scanned for any defects (using cameras and Xrays, etc), and then from there drive themselves to any given maintenance line appropriate for necessary servicing. Components can be rapidly removed and replaced, robotically.
Not only is this highly economical, but as a system it ensures the extreme reliability of the bulk of the vehicle fleet. We could have this kind of operation for maybe 4 totally standardised core vehicle types, especially including the commuter vehicles which would be the most common type of all.
Obviously there's a place and demand for special out-sized vehicles, of which will require more specialised (and slow) servicing, but for 99+% of the demand standardisation should be embraced.
Rail + Rubber?
With full-automation, you have cars that can drive themselves along the road with rail-line type precision. This provides for another interesting opportunity.
The idea is to employ a central rail wheel in between the rubber wheels of (some) of the cars - mostly freighters. The steel wheel extends and retracts as a load-bearing option only, for the sake of reducing rolling losses over longer distances. The steel wheel does not offer drive and braking, so it's simple. Drive and braking is achieved through the rubber wheels only, and the steel wheel extends and retracts so as to increase the traction of the rubber wheels as required in operation.
The system would require electronic stabilisation of the suspension of the rubber wheels (or it will tend to 'rock' about the pivot point that is the rail) but this should hardly be a problem.
This idea is not new. It has been investigated comprehensively by the Bladerunner development, as included (here).
The issue of sensor cost:
Some commentators have mentioned that they believe driverless cars will be a long time coming, due to the costly sensor/computational apparatus. Yes, they cost a lot for the prototype systems Google is using today, but there's every reason to believe that cost will drop progressively, and heavily.
Firstly, with driverless cars you move to network-based transport, which means 1 car will serve 20 or 30 people, which in turn makes otherwise costly technology affordable by default. Even still, maybe $100,000 worth of sensor technology per-car is a heavy bill to have to front-up with.
However, we can see that when economies-of-scale come into play, there will be no reason to believe that we cannot produce the supporting hardware for driverless cars at, theoretically, a fraction of that cost. To explain:
The reason why you can buy a small electronic system that is hugely complex, but at a very low price, is because of two things: It's mass-produced and it's small. With robotized production lines, the physical process of fabricating a given product (per-unit) accelerates in time the smaller it gets, as the physical processes of production speeds up. Putting it simply, it takes less time to drill a small hole than a large hole, etc. Anything small can, in theory, be built cheaply even if it's highly complex and precise, so long as there's a large enough production demand for it, likewise facilitating heavily investment in 'high-speed' automated production lines.
Sensor technology certainly comes under this category, like all information electronics. And indeed we have seen the predictable, progressive lowering of sensor costs over the years, which is a trend that should continue today and into the future - especially as ever-lower costs drive higher demand, which in turn drives more investment in production plants, and miniaturisation, etc. Also of course, the smaller a product is the less material it requires, compounding the reduction in production costs.
There are other factors which should drive down sensor costs. Google's current hardware system for its prototypes is highly elaborate, and probably an overkill. With refinement Google will probably find they can do away with a lot of their more expensive apparatus. But this is not their current focus, and nor should it be. Google's driverless car project is primarily in the business of software development (at this stage) which makes sense, as this is what the race-to-market will be essentially dictated by.
And after the "big push" from Google, and maybe others, there will be the "big pull". When governments realise driverless cars will be on the roads in a only few years time, and realise this conclusively, there should be investment in road-based sensors which take the pressure off driverless cars to do all the sensing and processing on their own. It makes sense for governments to invest in road-based systems in support of a fully optimised system.
To conclude, the current costly sensor technology is not really an issue, and will almost certainly become a non-issue, quickly, as time goes by.