Autonomous Vehicles – are they in the offing?

The advocates of driverless cars (autonomous vehicles, AVs) think they are already on the horizon and will reduce road deaths, pollution and traffic snarl ups. Their imminent arrival is seen as another disruptive influence on “current business” like  Blabla car share, Uber and AirbnB. These respectively aim to disrupt the status quo in medium distance and local transport and urban accommodation respectively. According to Paul Mason theses could be the harbingers of “dotcommunism”.

There is a key difference between these systems and AV’s. BlaBla et al are data based software systems that interact with the messy real world through people using Apps. AVs are software controlled hardware that interact with the real world by their sensors and software.

Piecemeal developments by individual motor manufacturers have led to vehicle features that seem to be the forerunners of this technology. Automatic gearboxes and cruise control systems have been around for decades, “park assist”, “lane control”, ”lane change” and “adaptive” cruise controls are becoming available. There are even subsystems to make sure the human driver is awake. So it might not seem unreasonable to assume that full autonomy would be a logical development from these systems.

But this is one of the many misconceptions about this technology. Current “driver assist systems” work on rule based “traditional” software. The complexity of driving in “the Real”* means that the software to operate these vehicle systems has to be more akin to Artificial Intelligence rather than complex computer logic.

Judging by the approach being taken by Google and others, AV software will be capable of learning from experience and taking actions based on it. One input to this learning should be the “rules of the road” of the local jurisprudence. Some governments believe that the manufacture and use of AV’s will offer significant economic advantages and are already putting in place regulations under which “on the road” testing may take place.  This regulation does not appear to include detailed test programmes or third party examination of the software systems that the recent furore over environmental testing indicates should be necessary.

So, AVs are on their way. The question is not so much when they’ll be here but how quickly they will be the only sort of vehicle in use. Their supporters paint a rosy future of adoption, because they believe everybody will want one for AV’s to take advantage of the time saving that their use will yield - the time currently “wasted” in driving. With an AV, you really can sleep as you drive to work, deliveries by AV will be quicker, for there’s no need for the driver to rest. Road accidents will be reduced because accidents are caused by the frailty of the human driver. AV’s will drive more closely to the vehicle in front, so can carry more traffic and, because the traffic will be a constant speed, AV’s will be more fuel efficient.

But an AV is expected to be more expensive than a conventional vehicle, so not everyone is going to be drawn in by such benefits. It would seem more realistic to assume there will be a mix of new and current technology where human drivers, pedestrians and cyclists will be mixing with AVs for some time. The mixing of humans and AVs is likely to cause trouble.

A human road user makes continuous observations to predict what others are likely to do, and communicates with them appropriately. This includes eye contact and exchanges of sign language – for example, to decide who should go first. Some situations require a human driver to have a loose interpretation of the “rules of the road” in order make progress e.g. when giving way to let someone out of a side road. Often, the decision to act is made quickly and almost without conscious thought.

This would not be a problem is AV’s were to be separated from human traffic – but that would require further investment in special “AV only” roads.

A lack of separation implies that an AV would have to include a code of ethics in its learning system; this leads to a question about who would take responsibility for its actions? Does the responsibility flow back to the manufacturer of the AV or to the person who last serviced it? A human driver probably has “self-preservation” pretty high in their personal “action oriented decision tree” which may override their personal ethics. Will an AV always have the same viewpoint, or will its view be different if it’s carrying live passengers rather than just “goods in transit”?

Can we have confidence that the AV programmers have thought that through and given their machines sufficient flexibility of action?

Assuming these issues can be fixed satisfactorily, there then is the issue of whether such cars are “closed systems”, or whether they would communicate with other AV road users and  traffic control systems – in fact both make eminent sense.

An AV communicating with traffic control systems should be able to  make better decisions about slowing down and stopping for (or even, perish the thought, “jumping”) traffic signals or clearing the way for public service and emergency vehicles. Communicating with other AV’s might enable the establishment of “road trains” which have been shown to reduce fuel consumption. They might even exchange real time road information with each other to warn of hazards.

There’s a downside to providing the AV’s with communication. Their systems could be vulnerable to attack by third parties.  This vulnerability, demonstrated on an existing vehicle, will require the software to be resistant to sophisticated “cyber” attacks. Governments might want access to AV system, particularly to the SatNav systems as this would enable viable road pricing schemes to be introduced. These are seen as an effective way to reduce pollution and further road development as well as being another attack on Civil liberty. However the authorities don’t need to know who is driving an AV – just who to bill.

AVs in operation will need to top up with fuel – if it really is driverless, will it have authority to charge the cost to its owner credit card or would it have its own account? Does a human need to own an AV? Their use would be a natural for an extension of an App like Uber to be summoned on demand.  Alternatively, an owner could use a combination of BlaBla, Airbnb and Uber Apps to rent one out to other people when it’s not required for her/his own use… 

We’re back to “dotcommunism”.

Finally, assuming these hurdles have technical and social solutions, how can the public be sure they are safe and fit for purpose? Given the recent discovery that a motor manufacturer has been “cheating” during emissions testing, governments have to think through how the vehicles will be tested and type approved. That could delay the whole project, particularly as there would have to be Government agreement at all levels.

AVs are not in the offing, they are beyond the horizon.

Notes

* A term (meaning the real world) borrowed from the “Culture” science fiction novels, , written by the late Ian M Banks

** This brings to mind the idea of the vehicle having a “personality” with a backup available in case of a “mental” break down – as exemplified by Eddie, the shipboard computer in Douglas Adams’ “The Hitchhikers guide to the Galaxy”.