Levelling the playing field for the driverless car: Part II

Catch up with Part I here>>>>

Standards for the ‘Right’ Choice?

The advent of automated driving may also lead to the requirement of standards for what could be described as ethical issues. One such potentially very problematic set of scenarios is being called the “death algorithm”. It is based on the assumption that, no matter how refined the technology will be, there will be crashes involving even the most advanced machines.

Self-driving cars will no doubt be equipped with systems that limit the damage, both to the car itself and to whatever and whomever it collides with.

Stephan Appt, a partner at the law firm Pinsent Masons Germany,poses the problem of a car facing an imminent crash and the its options being limited to swerving into a cyclist wearing a helmet or a cyclist not wearing a helmet. For the sake of crash-optimisation, a programmed self-driving car should probably crash into whatever can best survive the collision, in this case the cyclist wearing the helmet. However, this would create the contradictory, and slightly absurd, situation of “selecting” for injury a person following safety standards rather than one who does not.

Some might argue that programmers will eventually have the task of creating a system that would, for dead-end situations, foresee hitting an elderly man instead of an infant, on the grounds that a baby has more of life ahead of it than the old man. "But this could trigger the question of criminal law responsibility for the programmer, and it is also a most difficult ethical problem, which must be addressed with clear legal guidance, or standards, for programmers on how to design crash algorithms,” Appt says.

This standard would presumably also include striking a heavier vehicle that can better absorb the shock than a smaller vehicle. This could raise additional legal problems if, for example, the driver of an SUV is killed in a crash because the “death algorithm” directed the crashing car into her vehicle instead of into a motorcyclist. It may well be that the choices are so difficult to resolve that detailed information regarding the possible targets of a crash will not be provided to the automated vehicle or the programs will be designed to make choice on the basis of randomness where choices seem ethically inappropriate.

New-System Standards

Of course, legal and ethical issues are only part of the automated-car ecosystem that will require standards updates in the years to come. The technology itself will require ISO to deal with new systems that need to react much more quickly and carry far more data than is now the case. These standards are particularly important in “off-the-shelf” systems such as the CAN Bus, which enables systems to communicate with each other.

Aria Etemad of Volkswagen Group Research says these standardised systems save carmakers a great deal of time and money. “If there is no standard, I have to test it and it’s much more expensive,” he says. “So we use as many off-the-shelf systems as we can.”

The increased number of systems in the automated car and the increased speed at which they have to communicate with each other is making the CAN Bus gradually obsolete. “The CAN Bus will not be adequate as the car moves to higher automation levels,” Etemad says. “Even now, cars have two or three CAN Buses, where there only used to be one.”

As a result, new systems are being considered to replace it, such as the FlexRay communications bus, which is 10 times more rapid than the CAN Bus, and even using internet protocols. The IPs will be put in cars to enable them to speak with each other, Etemad notes. They could be used within the car as well.

“ISO needs to develop new standards for these new systems,” he says, adding that developing standards is a “complex and time-consuming process”.

As new systems are conceived and developed for the increasingly automated car, standards will be needed for the testing, validation and verification of these systems, Etemad says. “We need comprehensive test and type-approval standards. This will ensure that the vehicle is tested properly and, after type approval, can be sold on the market.” Currently, standards for testing and type approval differ slightly from country to country.

As the vehicles tested become more automated, standards will also be needed for test simulations, simply because the real-word testing of these complex systems will become far too costly, Etemad says.

“For testing new ADAS systems you need to drive approximately 1Mkms to make sure that the system works the way it is designed,” he recounts. “With an automated vehicle, you will need perhaps 50 to 100 times that – up to 100Mkms. That’s because it does not cover just a specific use case, such as adaptive cruise control, but it covers everything. Nobody can pay for that. You need simulation and perhaps you need standards that describe your test procedures.”

These standards could encompass measurability, evaluation and verification procedures, as well as the software used in the simulations.

Etemad says standards will also be needed for the hardware that provides continual real-time data to the car’s automated systems, such as sensors, the frequency of radar systems, laser systems and a standardised interface for cameras. “This would make things easier for all of us,” he declares.

Standards for a Complex Network

Regardless of how autonomous the autonomous car will ultimately be, as it becomes more of a computer and as the driver becomes more of a passenger, there will be a concomitant growing need for constant, bug-free ultra-high-speed connectivity. The self-driving car will need constant access to continuously updated maps and real-time traffic data, as well as information on road conditions and weather. And as the car becomes part of a multimodal routing system, it will also require data on parking availability, public transport schedules and even pollution levels.

In addition, as Appt points out, the failure of a mobile network could lead to an accident, either involving another car or pedestrians. With lives at stake, as well as the viability of complex urban and inter-urban road networks, regulators will need to provide clear standards so that, as Andrea Sroczynski, head of Market and Partner Strategy at Telenor Connexion, puts it, “cars and the ecosystem can communicate seamlessly”.

She says connectivity will not only involve communication among cars and between cars and the infrastructure but will also include streets, buses, motorcycles, cyclists, pedestrians and even dogs. “For this you must have strict V2X standards, either locally or regionally. You also need a more open ecosystem.”

Unfortunately, Sroczynski notes, the car is evolving more rapidly than the infrastructure, at least in Europe, because of the relative conservatism of public authorities in some countries. “This makes it tricky to put an autonomous car on the road,” she notes wryly.

In the United States, on the other hand, public authorities – notably the U.S. Department of Transportation (DOT) – have been actively involved in moving forward the development of a connected infrastructure. The DOT has initiated and funded a number of recent and ongoing research projects, carried out in collaboration with a large number of carmakers and Tier 1s, to speed up the development of V2V and V2X technology, car connectivity and, more recently, automated car systems and, in the process, define a set of standards for the technology.

In Europe, Sroczynski notes, this has been driven largely by carmakers and Tier1s, notably the Car 2 Car Communication Consortium, a broad association of carmakers and suppliers whose objectives include developing an open European standard for C-ITS, helping in the development of standards for V2I interoperability with the V2V standard and pushing for the allocation of a royalty-free pan-European frequency band for V2V applications.

Sroczynski has long been an advocate of an open platform approach to connectivity and this will become more important as more and more data needs to be communicated among all the players in the ecosystem. “Take as an example a cyclist on the road,” she says. “He will communicate his location over Bluetooth via the HMI of his phone. Then the back ends will need to communicate. Will there be a super-cloud where everybody can feed in? Or will it be one-to-one communication?”

If it is a cloud, the question then becomes who will run it. “It could be a public authority,” she suggests but adds: “Look what happened with eCall and all the delays in launching it.”

Players in the ecosystem are “getting more open now,” she notes, “which means that industry players could take the lead in this, perhaps as a consortium of carmakers, MNOs and public bodies.”

Standards will then be needed for the degree of openness in the cloud, Sroczynski says. “It will only be open to a certain degree,” she explains. “Part of it could be open for the super-public, some data will have to be shared among all the players and there will be ‘walled gardens’ restricted to, for example, Volvo owners or the users of a certain fitness tracking device.”

Finally, with so much data being generated and communicated via mobile networks, setting security standards will become of vital importance, Sroczynski says. “MNOs are good already in security but now we will be dealing with life-critical systems and this requires a dedicated setup such as we have at Telenor.”

Catch up with the latest thinking at Consumer Telematics Show 2016 January 5.


Leave a comment

Your email address will not be published. Required fields are marked *