The Importance of a thorough design validation regime in the development of Autonomous Road Transport

Many technical advances are afoot in the autonomous road vehicle domain and, in some senses, ‘driverless cars’ are already amongst us in the world. However, I believe that great care needs to be taken in introducing this exciting new technology at large, which offers so much potential in improving the quality and safety of ‘end-to-end’ journeys.

The challenges in introducing this new technology in the safest and most beneficial way are dependent on both technology and societal issues. Here, I will focus on the technology because that is my area of expertise, but those social issues will be ignored at our peril. In other words, quite apart from the relative immaturity of the technology itself, I do not think society, at large, is ready for the sudden introduction of fully autonomous vehicles, we are much better becoming accustomed to them as the new technology is introduced gradually.

Driving along motorway style highways is, from many perspectives, the ‘low hanging fruit’ of the overall challenge. Here, most vehicles are going in the same direction at much the same speed, with clear rules of ‘engagement’ between vehicles, away from distractions and dangers such as pedestrians and cyclists.

Of course, this is often one of the situations where this new technology could be very useful, since we are often behaving pretty closely to “automata” in these circumstances and we are not particularly well suited to such situations in terms of maintaining our awareness. In this situation, a multi-sensor ‘all round view’ autonomous car with reaction times that could typically be at least ten times faster than a human’s could save a great deal of boredom but also, more importantly, injury and even lives, because it will never get bored and will have ‘eyes in the back of its head’.

However, although we have become accustomed to the fact that car travel is a really very unsafe way to get around in general, motorways are relatively safe. Inner city areas see a larger number of accidents, though they tend to be less serious, and the real ‘danger spot’ is the relatively high-speed two- or three-lane suburban/rural highway, the ring roads and ‘A’ trunk roads of the UK. Here, relative speeds between vehicles can still be high, and it is a much more complicated and less structured environment from both a sensor and control perspective. It is inner city and suburban areas where the VENTURER project [2] focuses, on both the required technology advances and human acceptance issues.

Since the technology is really not there yet for these types of environments, there is a great deal of research, and validation of the outcomes of that research, to be carried out over the coming few years. In fact, one could also say that even on motorway driving scenarios a great deal more technical validation and test-track proving for unusual situations is still urgently needed before letting this technology loose on public roads in earnest.

Let’s consider just two issues worthy of brief note here, in order to give a feeling for the scale of challenges to be dealt with, are as follows:

  • Planned control handover: let us assume that you want to drive your car to the motorway junction, let it do the ‘boring bit’ of going up the motorway, then warn you that you should take back control to drive on the smaller roads to your destination. What happens if, by that time, you are asleep, or immersed in reading a book? How should the Autonomous Control System (ACS) handle giving control back to you and ascertain whether you are in an appropriate attentional state to do so? Emergency handover is even more contentious, and it is hard to think of many times when this would be a good idea, compared to the ACS finding a safe way to bring the journey to a temporary halt.
  • Turning out from a minor-road T-junction onto the main highway: it is a busy main road, and there could never in the foreseeable reason future, be an allowable chanced for the manoeuvre whilst obeying the highway code. If driven manually, then the driver would find ways, eventually, to signal a request to approaching traffic using a wide variety of methods. How would an autonomous vehicle do this, especially if communicating with another manually driven vehicle?

These two examples are the first two (of three) trial topics for the VENTURER project, which will consider the technology and its impact.

Professor Tony Pipe, Bristol Robotics Laboratory [1], UWE Bristol[2]

[1] The author is Deputy Director of the Bristol Robotics Laboratory http://www.brl.ac.uk/

[2] The author is leading technology integration in the Innovate UK-funded autonomous vehicles project ‘Venturer’ https://www.venturer-cars.com/.