Are we ready for the self-driving car?
Thursday 4 February, 12.30 – 1.30pm - session now closed
Hosted by Tyron Louw, PhD student, Institute for Transport Studies, University of Leeds
Thanks to Tyron for hosting this excellent session, and for providing such comprehensive and thought-provoking responses. If you would still like to put a question to Tyron please use the box at the foot of the page and he will respond, but not in 'real time'!
This live online forum looked at the issue of autonomous vehicles, and whether road users in the UK are ready to embrace the technology. The forum was open to anyone with no need to subscribe or log in.
Tyron Louw's main research area is in driver behaviour and his PhD research focuses on the human factors of highly automated driving. He is specifically interested in how to safely re-engage the driver in manual driving.
Tyron is also a research assistant on the EU funded projects CityMobil2 and AdaptIVe, where he investigates road users' interactions with automated road transport systems and tries to understand how drivers can safely interact with automated vehicles.
To submit a question please use the box at the foot of this page.
Q1: POSTED BY: MATT STATON
Most of the trials/research into driverless vehicles that I have read about, and the accompanying legislation, focus on the role of an "operator" or "driver" being in the vehicle to take control in the event of an emergency. If we are to truly see "driverless" vehicles on our roads, at what point do you expect a shift to the human in the vehicle being referred to as a "passenger" and what barriers do you see there are to that happening?
Estimates tend to vary. Some say 10 years, others say 40, others say never. I think the point at which our vehicles will be so intelligent that they can traverse all manner of road, weather and traffic scenario (which unpredictable human drivers are) is still a long way off, which means that humans will still be involved on some level for some time.
There are a couple of barriers, but I think one of the main ones will be the technology. Computers do everything you tell them to do (mostly) but nothing that you don't, so as long as human drivers are on the road and there is some element of unpredictability on the route or in the environment then you can expect that humans will still need to be involved. Of course, here we are not talking about segregated areas for automated vehicles. We are talking about normal highways/urban areas where the vehicles would have to interact with other road users, including other (human) drivers, buses, pedestrians, cyclists, and in some places, goats, tuk-tuks, cows, and even wildlife!
I also suspect that, despite introducing ever higher levels of vehicle automation, car makers have some vested interest in having their driver see themselves as “drivers” and not only as “passengers” (otherwise, they’d be directly competing with the likes of Google, which I am not sure they are ready for just yet). In fact, only the other day Porsche CEO Oliver Blume said that he is not at all interested in bringing driverless technology to Porsche, commenting, "An iPhone belongs in your pocket, not on the road". That’s an extreme example, and it's not necessarily a barrier to full-vehicle automation, but the big car makers are going to be trying hard to stay relevant in our futures.
Q2: POSTED BY: DUNCAN MACKILLOP
The revelation that the human drivers of the Google cars have had to intervene 'thousands' of times leads us to suppose that there is often a mismatch between the car's probability/possibility trade off and the driver's. This type of situation where the car is happy, but the driver isn't could lead to all sorts of problems with the acceptance of the technology. My question is therefore how does the industry propose to address this problem?
Though disengagements are going to be a reality in the future, it is important to note that those listed in the disengagement reports were done during testing (and many in simulations). In our lab, we also observe driver initiated disengagements and are working with car companies in the EU-funded project AdaptIVe (https://www.adaptive-ip.eu/) to understand how drivers behave while they are in a highly automated vehicle, what makes them uncomfortable, and how they respond when they are uncomfortable.
I can’t speak for carmakers, but they will understand that the acceptance of this technology will make or break them in 'Car Industry 2.0', so they are going to do everything they can to understand what their consumer base is expecting, and then deliver on that. (For instance, you will see the philosophies of the different companies coming through in the way that they market their automated technologies. Be sure that Volvo and BMW are going to have different approaches). One strategy to ease the public into it is to continue the trend of gradually introducing the technology to the market with successive models. The chances of misuse, abuse, disuse are real and we’ve observed some instances of these already in videos on YouTube, and I guess the way they will address the problem is by doing extensive research both privately and in integrated projects like the one I mentioned above.
Q3: POSTED BY: ROB WILTSHER
Where do you stand legally if the car is involved in a collision? Is it the car's fault or yours?
I am not a legal expert but I strongly suspect that the liability will fall on the carmakers. In what I am sure is going to be the market trend, Volvo recently announced that it will assume responsibility if it's automated vehicle causes a collision.
This is in line with our thinking from a human factors perspective, as we know that humans are poor supervisors, and we can’t reasonably rely on humans who are out of the loop. The following quote sums up the irony of vehicle automation: "If you build vehicles where humans are rarely required to respond, then they will rarely respond when required"- Peter Hancock
Q4: POSTED BY: MATT STATON
When do you think we will see driverless vehicles on public roads in the UK (if at all)?
The Government has made it clear that it wants to attract vehicle manufacturers and others to the UK to test their automated vehicles. It's not unlikely that in the next few years we will see, for example, sections of motorways dedicated to conducting this testing. As an example of what is possible you can have a look at Volvo's DriveMe project, which will involve 100 highly automated vehicles (CX90) operating in Gothenburg in 2017 (http://www.volvocars.com/intl/about/our-innovation-brands/intellisafe/intellisafe-autopilot/drive-me). It is difficult to predict exactly when or in what form this will happen in the UK but if the Government’s appetite is real then I think it will be sooner rather than later.
Q5: POSTED BY: MATT STATON
Do you think the introduction of driverless cars will replace the use of current forms of public transport i.e. bus, taxi, tram or train?
I very much doubt that. I can see how UBER self-driving taxis could replace human-driven taxis in the future, but mass public transport is unlikely to be replaced by small driverless vehicles for one main reason: capacity. It’s another thing entirely to ask if public transport will be automated. Indeed, there are examples around the world of this already happening, albeit mostly in segregated areas, otherwise they will be at a very low-speed. Examples of these include Docklands Light Railway, the pods at Heathrow Terminal 5, rail system at Charles de Gaulle Airport, and the automated buses in the EU-funded project CityMobil2, which are being demonstrated at sites in various cities across Europe. In the future, we are likely to see these forms of transport become more automated, but I don’t think they will be replaced by limited capacity driverless cars.
Q6: POSTED BY: PAUL GANNON
What guarantees can be offered that 1) users cannot insert software that overrides safety parameters & 2) that external parties ('hackers') cannot insert software that overrides safety parameters?
I think at this early stage few would want to offer this guarantee. I do know that security forms a substantial part of any significant discussion on vehicle automation. Concerning customisation if carmakers are assuming liability if something happens to you in their automated vehicle I don’t think that they will give too much room for users to set their safety parameters. Therefore, I think they will be working hard at making their system difficult to ‘hack’.
Q7: POSTED BY: PAUL GANNON
A recent newspaper report from the US said manufacturers & police are concerned that collisions are occurring because automatic cars are driving carefully and software may need to make them drive less safely. Also many users will not want to be limited by safe driving & will attempt to override safety parameters. How likely is it that automatic cars will be designed to drive unsafely like human drivers?
I think it’s going to be difficult for carmakers to programme their automated vehicles to break the law. The only hope, I think, is that human drivers become less emotionally attached to the act of driving and accept the trade off. Most people, when they get on a bus, aren’t obsessing over their journey time. They know when the bus is leaving and when they need to get to their destination, and then plan accordingly. A few erratic or disorganised individuals shouldn’t mean that automated vehicles should drive unsafely. I think people who will not want to be limited by safe driving will end up voting with their wallets. Automated vehicles are coming, but they won’t be mandatory!
Q8: POSTED BY: PETER WILSON
How long do you think it will take an operator to a) take physical control when told by the vehicle and b) get into driving mode. Will these 2 things increase the potential for human error incidents?
Driver’s can resume control in 1-2secs if the alarm is loud enough and the driver knows that the alarm means that they need to take over control. However, to answer point b, it's not simply a case of taking back control. How long it takes drivers to get into “driving mode” depends entirely on the situations that they will have to deal with when they resume control. If they simply have to resume control and keep the car in a straight line, then they’ll be able to do that relatively quickly with no hassle. However, if they have to resume control because its foggy and there is construction work ahead that they have to avoid then it may take them a bit longer to get into “driving mode”. When you’ve not been driving for few hours, it's not only about waking up and understanding where you are and what you need to do (regaining situation awareness), but also having to re-learn what steering/braking force results in what vehicle outputs. You are correct, the more demanding the scenario, the greater the chance that something could go wrong.
Q9: POSTED BY: JAMES GIBSON
Do you think cars will ever be fully automated or will there always be the need for a driver control of some kind to make the car move, be it a steering wheel, joystick, or buttons? I ask this because how will a fully automated car cope with for example the final detailed part of a journey? Currently Sat Nav technology often takes you close to your location but not down to the exact place you want to be. If I think of my drive at home, how could you programme the car to be 2.5 metres from the front of the house and 1 metre away from the hedge so that I can get my wheelie bin past ready for collection?
Technically speaking if a vehicle can’t go from A to B and perform all functions in-between (including negotiating the wheelie bin) then it can’t be considered fully automated. When we hear the term “fully self-driving vehicle” it will often refer to a vehicle, or ‘pods’ in the case of the UK demonstrations, operating in a somewhat controlled and ‘known’ environment. That said, I don’t think the parking aspect of the technology will be too difficult to overcome. The low speeds make it much easier, especially if there is a system where you can teach your vehicle how you want it to park. Indeed, Tesla and others have recently released videos showing off how their vehicles can park themselves at home, but we’re yet to see how they will handle unknown destinations. To be fair to the technology, if we don’t quite know where we want to end up I don’t think we can expect our cars to! I think over the next few decades we will see the human being weaned off the major parts of driving, but I wouldn’t say that we’ll never be involved.
Q10: POSTED BY: MIKE YEOMANS
Do you think we might phase the acceptance of automation, by introducing into current vehicles adaptations that we might become familiar with prior to the big switch over, no doubt some years from now? For example, we already see autonomous braking, parking assist etc., currently as safety features when the driver fails, is this now leading to the features taking over? Today's driver, once comfortable with these features, will no doubt begin to rely on them - could this be intentional as a phasing-in of confidence in autonomous vehicles.
As I mentioned in a previous question, I think that for carmakers this will be the natural way of introducing driverless technology. I think there are two reasons why car companies would employ this approach. The first is a technology acceptance issue, as you rightly suggest. They know that a fast transition to a radically different driving experience will be a big ask for many people, and the car companies need to feel their way into the future carefully so as to stay relevant and keep their consumer base. The second is that they want to remain financially viable. Many car companies are sitting on technologies and features now that they only plan on releasing in a few years (or models) time. At the moment the technology is expensive and they will be careful to not price themselves out of the market. On the other hand, I think companies like Google, UBER, and even possibly Apple (if you believe the rumours) are less concerned about the financial viability at this stage. They are focused on getting the technology working, and by working I mean they are aiming for the type of driverless technology that would see humans be passengers more than drivers.
Q11: POSTED BY NICK HUGHES
How reliant (if at all) are driverless cars on road markings? I ask this in the context of many markings not being refreshed until they are almost worn away and considering that some roads have had the centre lines removed with more to follow most likely.
I am not an expert on the sensors, but I do know that the technology designed to track road markings is only as good as the quality of the road markings. Indeed, there are examples online of systems disengaging for this very reason, where the driver has to resume control. However, it's unlikely that driverless cars (with that level of responsibility) will only have one level of 'vision' as it were. Most systems will integrate LIDAR/RADAR, camera-based systems, GPS, etc. However, it is important to note that this is only one model of driverless car technology. The ‘Googles’ of the world are pushing to develop this type of system, one that can operate independently in any setting. However, there is another model, connected automation, where vehicle-to-vehicle communication and vehicle-to-infrastructure communication provides another level of support and 'vision'. For example, the car ahead of you telling your car that the car ahead of it is braking. The second model is possibly more reliable in the sense that things like poor road markings won't necessarily cause a disengagement, but it also requires a lot more infrastructure and planning than the first model. Another real challenge to the second model is data sharing: how likely is it that car companies will be happy to share data - in real time - with anyone in the vicinity? I'm not so sure!
Q12: POSTED BY: HUGH JONES
With the over-reliance and over-usage over here (but hardly at all in the US) of headlamp flashing (wrongly) used to 'give way'; acknowledge; give thanks; etc. how will a driverless car know how to interpret or ignore these 'signals' as and when required, without causing problems?
The cars will first have to be programmed to recognise and interpret the signals - it won’t need to ignore a signal if it doesn’t register it (as humans we don’t have the option of not registering sensory information). In ambiguous situations, and not at, for example, traffic lights where the rules and rights of way are clear, I imagine that the vehicles would either be cautious and give way to the other vehicle or hand control back to the driver. Carmakers are slowly starting to think of ways that automated vehicles can communicate their intentions and messages to other road users, including pedestrians and cyclists. If they get this right it will allow automated vehicles to operate in a mixed traffic environment more naturally.
Q13: POSTED BY: NICK HUGHES
I would expect that even with fully autonomous vehicles there would still be some collisions / casualties on a public highway network. Do you think that the public would accept a lower level of casualties or would they/we demand such systems to eradicate all casualties before it would have public acceptance to its adoption? I personally find it difficult to accept that all pedestrian casualties would be prevented due to an apparent desire by some humans to enter a carriageway in front of a vehicles within its stopping distance? Humans trade off the convenience of air travel against a low risk of being involved in a plane crash so might the same be true on the ground?
There will, of course, be collisions and casualties. Human error plays some role in the majority of accidents. To eliminate that completely you need to remove the human altogether, however given that there will be a considerable period where humans and automated vehicles share the road, the likelihood of human error still cropping up is high. Indeed, we have identified new types of human error that have not previously been an issue in driving but have been observed in supervisory tasks, such as mode confusion, overreliance on automation, vigilance decrement from being a passive monitor of the system. There is a lot of talk about the 'first crash' (caused by an automated vehicle), especially in the media, and it remains to be seen how the public will respond to that. That said, I think most people will come to accept the trade off. Yes, some choose to not fly to avoid the risk of being involved in a plane crash, but that's not the majority. One obviously cannot control all elements in the environment, but if avoiding a collision with a pedestrian comes down to perceiving them early enough and reacting to them quick enough then I'll take a bet that an advanced automated driving system would do it better than me, especially if my head is in the clouds and I'm not expecting it. The success of these systems will be judged on their overall impact. It's natural to think of how it might impact us individually in the worst case scenario but governments and industry will push it so long as accidents and deaths on the road are reduced on the whole. As unfortunate as it is, we should be prepared for at least some collateral damage in the pursuit of the sort of thing Volvo is aiming for in their Vision 2020 (http://www.volvocars.com/intl/about/our-stories/made-by-sweden/vision-2020). "Our vision is that by 2020 no one should be killed or seriously injured in a new Volvo car." - Håkan Samuelsson, President and CEO, Volvo Cars
Q14: POSTED BY: HUGH JONES
Despite the fact that trains travel on rails, on fixed courses and with little or no need to vary speed and with no interaction with other moving things and could therefore be said to be the least risky and complicated means of travel, a human is still required to be at the controls. This being the case, how can we expect road going motor vehicles ever to not need a human at the controls?
It's a good point and it will be an enduring challenge for vehicle automation. We may see sections of the road where vehicle automation can function without human intervention, for example, on highways that are connected. The point you make lends strong support to the argument that the full benefit and use of automated vehicles will only be realised in totally controlled environments and routes, which somewhat diminishes their wider value.
Q15: POSTED BY: ROB BESWICK
Do you anticipate that people in control of driver-less/automated cars would still need to pass a traditional driving test? Do you foresee a time when people will be in control of such cars without taking a formal qualification in driving?
As long as humans are involved on some level I think they will always need to have a licence, even if car makers assume responsibility in some situations. It's not unreasonable to think that the traditional driving test will change as the nature of the driving task changes, but I think it's more likely that individual car makers will have vehicle/system-specific training. I say this because at the moment there are no international standards for the design and technical configuration of the systems that would automate a vehicle, and while every effort is being made to design intuitive systems there are some things that still need a manual to do! Therefore, I think carmakers would be best placed to deliver the training, and if they are assuming responsibility it would be in their best interest to make sure their drivers understand how to use the systems they have bought.