Autonomous Vehicles are Not New
In the last year, unmanned cars have made the headlines throughout the world, and the idea that cars could be driven solely by computers came as a shock to many. The reality is that driverless vehicles have been around for some time, both in railway and PRT operations; and analyzing how driverless technology has been implemented there, should help understand how this technology will be brought to market and the hurdles that the automotive industry will face.
Taking into consideration key success factors from the two industries with driverless experience, we foresee a continuum of new technologies, which will be brought to market, before any car can show real autonomy. Though some manufacturers and car specialists like to picture driverless cars as completely standalone solutions, they will never be autonomous or independent, as driven cars are today. The reason is quite simple: they will rely on vehicle to vehicle (V2V) and vehicle to infrastructure (V2I) communication.
It is obviously less exciting to picture a system approach rather than a solution mimicking today’s reality where all intelligence is onboard, in the driver’s brain. Skeptics to a system approach will argue that the fleet of 23 Lexus Google cars that didn’t dispose of V2V or V2I capacities were able to run more than 1.7 million miles. That’s true, but they did suffer 11 accidents.
Even though none of these accidents could be imputable to Google’s driverless technology but human errors of other drivers, if you believe like I do that 100% accident-free must be the norm for driverless cars, this isn’t good enough. To be more accurate, the industry will need to meet the stringent requirements of 10-9 failure rate for critical elements as it is the standard for driverless trains.
The point is that the only way to avoid systematically rear end or lateral collisions is to get fore or immediate knowledge of deceleration from the car in front or from a car changing lane and to apply safety principles guiding all movements.
Vehicular Ad Hoc Network (VANET)
VANET technologies with proper software should take care of communication requirements. For safety principles, the automotive industry can look to the railway industry, which has been dealing with safety issues for more than 200 years. Railway engineers have come up with four general principles and thousands of technologies that ensure that these principles are systematically obeyed:
- Block interlocking: under normal circumstances, sections of railway – known as blocks – can allow only one vehicle at a time. Thus, vehicles cannot collide with each other if they are not permitted to occupy simultaneously the same section. However, this is only a certainty if vehicles are spaced far enough apart to ensure that they cannot collide, taking into consideration factors such as speed, braking time, and ramp.
- Block signaling: signaling to the next vehicle (with driver or unmanned) that no vehicle is in that section.
- Integrity: Even though a vehicle might have left a block section, there is always the possibility that a trailer could have separated from the head vehicle, which is especially crucial for long freight convoys. Beside this “carriage integrity” principle, which will be crucial to solve if platooning technology is to be successful, railways also have the track integrity issue, which will also exist for cars (potholes, rocks).
- Interoperability: Though the decision to make a system interoperable isn’t taken for safety reasons but rather for political vision or economic choices, it has safety implications. Once the decision is made to integrate various elements of a system, these elements need to work together safely. Physical but also conceptual interoperability needs to be sought to integrate various networks.
Lessons to be Learned
The automotive industry will need to make these principles its own with a few adaptations, creating new technologies and a multi-billion dollar market by the same token. Car manufacturers will also have to rely on international standards with stringent homologation processes – something they have not been good at in the past.