Currently engineers & scientists building automated self-driving cars, are attempting two methods:
- Easier (actually very difficult):
Cars that sense the road, and assist human drivers for specific tasks like self-parking, or avoiding accidents, or highway driving. And Elon Musk says Tesla’s high-tech cars will run on ‘auto-pilot’ 90% of the time by 2015.
These are useful features for a human driver to have, but they don’t create any of the incredible systemic changes & benefits of true self-driving ‘robot’ cars.
- Harder (even more difficult):
Its the last 10% of true self-driving, that Google is attempting — truly autonomous cars, that drive themselves anywhere by sensing everything. They dont even need to provide steering wheels…
In both techniques, the car has sensors like GPS mapping, and laser LIDAR ‘eyes’ that detect everything from pedestrians to obstacles and other cars.
Autonomous driving ability is hugely difficult to get right, for engineering, real-world, and even insurance & legal liability reasons, by current laws.
This video shows just what Google is trying to build and overcome. You can see here how they’re attempting to make a car ‘drive like a human’ so it signals to human drivers in familiar ways. The human ‘safety driver’ testing the vehicle also talks about the automated cars ‘assertiveness’ as a driver.
Even if perfected, Googles system might work in western countries with mature traffic rules, systems and human behaviours…but it would likely be challenged in the less organised roads of the world, from Africa to India, as seen here.
The dream of total automation could prove surprisingly elusive.
MIT Technology Review:‘Driverless Cars are further away than you think’
The key factor is human behaviour — unpredictable actions for a computer system that counts on knowing what comes next to a high degree. Not to mention the risk and dilemma of Insurance claims for damage, based on decisions made by a computer – who is liable? the owner of the automated car? the company that made it? The company that wrote the software? The company that made the cars sensors?
Take the trouble with diagnosing why your laptop may be malfunctioning – is it something you did, an app, the operating system, the hardware, some hidden setting maybe? – and multiply that by a factor of life and death.
Beyond engineering, for truly automated cars, at the minimum there are issues of:
- making it affordable given the current sales model
- putting enough on the roads for the system to not be a niche technology and really have network benefits
- old road safety laws and regulations that are made for a human-driver context
- Insurance assumptions that are made for a human-driver context
But why try and make a car drive ‘like a human’?
Its a transportation system, not a person.
So, do Automated cars have to be Independent?
That is not a trick question. Automated transportation, and cars that drive themselves (though very cool) – are two very different things.
Could there be a middle ground to current methods? Something that attempts to simplify the engineering and cost issues, while addressing the legal ones?
The Idea: Automated Cars, but following painted ‘rail-tracks’ on the road
These ‘Smart’ stripes, painted onto roads, would be:
- Easy to paint onto existing roads with current equipment and manageable budgets.
- Stripes would have something like RFID chips embedded every short distance. These passive digital markers can be useful for storing basic data about the road/terrain/last 5 cars ahead and their speed. This could be useful to supplement any cars data/information passively for added safety, if wireless network signals fail or get blocked.
- Can be maintained as a part of regular public city infrastructure, and provides an interface between public city systems, and private vehicles, thus reducing the burden on a cars system to understand and keep up with any change in road conditions/situations. Even at an immediate local level in case of accidents, police calls in a neighbourhood, Public services can ‘write’ information into the digital cloud, as well as the physical road itself, as backup.
- Stripes are easy for pedestrians, bicyclists and non-autonomous cars to see, keep an eye on, or avoid, and include within existing signalling regulations.
- Stripes could also mark specific road lanes, for autonomous cars alone – from highways, to suburban streets.
- Stripes can be extended from the road right into parking spots by building owners themselves – making it fast to implement by dividing the work. Imagine knowing where the free spots are when you drive in, and being able to pick or pay for one in advance.
Imagine city roads painted like a giant circuit board with smart strips — a passive, low-tech, backup system to gps and wireless data
- As I understand it, if these stripes are implemented, Car sensors will need to mostly just detect obstacles well, and talk to other cars nearby about their intentions like turns, speeds, and heading. They can also leave their motion data on the RFID in the Stripes for following cars as a failsafe.
- Since the road ‘speaks’ to cars through the stripes, its a backup system to tracking their ‘real’ location on the road via GPS.
Its then less critical, though still handy for route/congestion planning etc.
- Stripes with RFID are also a level of failsafe in case of failure of the wireless or GPS signal, or an onboard computer crash. ‘Dumber’ and less complex sensors can use the road-stripe data to safely come to a halt, if the main computer stops sending signals, or sends conflicting messages.
I’m sure there are issues I haven’t thought of with this system, but it seems to hit the spots in my mind.
- Simplicity – Reduces the complexity of Independent Automated Cars (that i can see with my limited knowledge)
- Sharing the burden of operation from the cars’ on-board system and ‘cloud’ with public city infrastructure
- Standardisation – Stripes can be used by any vehicle made by any company in a standard way
- Vehicle types – Works for Trucks, cars, golf-carts and motorcycles in the same way, for local or long distance travel.
- Encourages governments and municipalities to get involved for the benefits, spreading the cost of rolling out true automated transport to the public.