The consequences of a migration to autonomous driving, as discussed in this opinion piece on autonomous vehicles by Dave Marples and Paul van Koningsbruggen.
Article from Thinking Highways, October 2015 - also available in pdf
Personal mobility, across its many forms, is changing, make no mistake about it. Improved Public Transport, Intelligent Travel Assistants and trackers are all optimizing the multimodal travel experience but they are as nothing compared to the changes we see in the world of the personal automobile.
Forecast automotive changes, both from within the industry and imposed from outside, include increased connectivity, the transition to leasing, novel fuels, vehicle tracking and Pay Per Use Insurance, environmental awareness (both in terms of emissions and in terms of vehicle lifecycle), security and integration of the vehicle with smart road infrastructure. These pressures all combine to create a future highway environment considerably different to the one we have today. These all pale into insignificance in comparison with the changes promised by Advanced Driver Assistance Systems (ADAS) and, even more so, by Autonomous Driving (AD) and Autonomously Driven Vehicles (ADV).
For the purpose of this article we consider Autonomous Driving to mean full-time automatic driving, with no requirement for human intervention at all, under any circumstances. By this definition there is no need for the user of the vehicle to have any special qualification and AD offers the fantastic promise of opening up personal transport to user groups who could previously not access it; The elderly, infirm, children and those suffering from a driving-affecting disability such as epilepsy or visual impairment. This alone fundamentally changes the role of the automobile in society. Just as an aside we have some difficulty getting our heads around the systems that some are proposing where control is given back to the human ‘driver’ when things get a bit ‘hairy’. It seems a little bit strange that the Autonomous System hands back control of the vehicle when there’s likely to be an incident…and who exactly is to blame in such a scenario anyway? It does, however, seem reasonable that a vehicle could have an AD ‘mode’ and a conventional ‘mode’ which are invoked at different times, but, please, not an AD mode that climbs out of the vehicle when the going gets tough! So, if we believe in true autonomy (and yes, it’ll take a lot longer to arrive than the popular press would lead you to believe) then we question what assumptions about the way our road networks operate become invalid, which ones will we miss and which ones are such natural transitions that we don’t even question them (ever noticed you don’t get dialtone on a mobile phone)? We decided to stare at the wall for a bit and see if we could start off this thought experiment – not least because there are changes we should probably be considering making now, if we truly believe in an ADV future. Most commentators have focused on the changes that Autonomous Driving will deliver for the driver and their passengers: increased freedom to do other things, the possibility of lower (or non-existent) license constraints on vehicle ‘conductors’ – we can’t call them ‘drivers’ any more – and improved safety under edge conditions. These are the most frequently quoted benefits, but a question that is just as interesting is what happens to the road network and society in this utopian model?
Society and Infrastructure
Some changes are obvious; one example of that is the nature of taxi rental changes from the rental of a person with a vehicle for a specific journey to the rental of the vehicle itself which, one hopes, will turn up at your feet on demand. Telling the vehicle where to pick you up might get interesting though – if you’re in a pedestrianised town centre you can’t ask it to arrive at your feet, you’re going to need a rendezvous point, so it’s reasonable to expect ‘mini ranks’ to appear where you can meet your ride…the current issues with Uber will be as nothing when AD-taxis start pulling into ranks! Enforcement of waiting restrictions will need to be automated and the allocation of fines will be difficult, so expect to be paying for your vehicle from the time that you agree it will be available to increase the likelihood of you being there on time and reducing the ‘dwell time’ at the rank.
While we’re on the subject of taxis and booking rides – why exactly do you need to park your car in town? Friends of the Earth inform us1 that the UK Department for Transport believes that two thirds of all car journeys are less than five miles (20 per cent are less than one mile!) so, when your vehicle has dropped you off, why not just send it home again to wait for a call to pick you up? In the timeframe we’re talking about no vehicle will be doing less than 70-80 MPG so we’re talking about 70p (€0.90) in fuel costs (today’s prices) to do that round trip…there aren’t too many places where you can park for less than a Euro. Of course, that impacts the number of vehicles on the roads, but not necessarily in a negative way. Some estimates place the number of vehicles in a city that are looking for a parking space at 30 per cent – so that traffic goes away, replaced by traffic that is heading into or out of the urban space directly; perhaps we’ll see Park-and-Rides replaced by Park-and-Waits.
There are open questions here though about the role of mass transit; municipalities are getting better and better at providing competent mass transit for access to urban centers. It’s quite possible that this will have been so well accepted by commuters in the timescale under consideration that they already simply won’t have vehicles; In the US we already see cities with a surfeit of car parking space (and suffering financially as a consequence) and we will see that trend accelerate due to AD. Some of that space can be given over to parks, pedestrianisation and other mixed use, but there are open questions of how to deal with multi-storey car parks which are difficult to re-purpose and are already viewed as an unpleasant, insecure, eyesore on our cityscape by the majority of residents; Hartford, CT considers that the surfeit of parking space it has due to the end of America’s love affair with the automobile already costs it US$50m a year in lost tax revenues2 – that’s a drop in the ocean compared to the future consequence.
For the present we need the majority of those parking spaces – after all, the transition to ADAS has hardly built up steam yet, never mind the transition to AD, but this is one of those areas where we need to start considering the future now – mass transit improvements and AD mean we won’t be making any long-term investments in car park companies! Furthermore, with the half-life of buildings somewhere in excess of 50 years our town planners should already be looking at how easily parking can be repurposed when demand drops – there are already some forward-looking cities, such as the Danish capital Copenhagen and Haarlem, Netherlands, already starting to do that with mixed use, repurposable, spaces in the design mix.
Let’s change gear slightly and look at how we might use the car. We already touched on the idea of sending the car home again rather than having it park up, but the increased user community also has an impact. Kids who hate walking to school could now take the car; a family that previously needed two cars because one was away with one partner at work might now get along with only one, albeit more heavily used. The reasons for getting a taxi (even one of these new-fangled AD ones) back from town after a night out are considerably reduced – after all, why not just hail your own vehicle?
The nature of our delivery companies is also hugely changed. No longer is a single driver responsible for a few hundred packages a day but now vehicles can be sent out, without drivers, to optimize the delivery process and return to base for more missions – Amazon, DHL and a few others are already hinting in this direction for drone deliveries, but an ADV is really just a land-locked drone. We’d like those vehicles to operate with a lower priority in the spaces between the regular, human carrying, traffic though, and that comes down to Ground Traffic Control, which we’ll deal with shortly.
What is the net effect of all of these behavioral changes on the support infrastructure needed for vehicles? How many vehicles will be transiting the urban environment at any point in time? What is the distinction between day traffic and night traffic in an environment where no driver needs to have got out of bed? It’s too difficult to say yet and modeling is needed –there’s a Ph.D. or two in there somewhere.
Changes to Vehicles and their support
Now, let’s take a step back and consider the behavior of the vehicle itself. Human beings are notoriously inventive at getting themselves out of awkward situations – vehicles simultaneously arriving at a roundabout, a car driving the wrong way up a one-way street or an unexpected hole in the ground rarely present an insurmountable obstacle for a determined driver, but its likely that there will be scenarios beyond the ken of any particular ADV. What happens then? We’re already seeing reports of ADVs not behaving in the way that other drivers expect, because they are accurately following the rules of the road, and with such rigidity it’s difficult to believe that the ADV will always be able to get out of every single situation, so the nature of Auto Rescue changes. We can’t allow just anyone to be able to CTRL-ALT-DELETE a vehicle to recover it out of a nasty situation (that would be a security nightmare) so some exception or escalation process needs to be created to recover the dumb mass that’s quite literally gone down a dead end.
Of course, for the vehicle, it’s not all bad. If we assume connectivity (and the new vehicle population is pretty much all connected already) then we gain some distinct benefits – we can control vehicles much more like we control airplanes. We can offer Ground Traffic Control, analogous to Air Traffic Control. We can instruct vehicles on how to transit a city and which route to take; if they are permitted to use range extender motors in this locality; what maximum (and minimum) speeds to adopt; what issues exist on the road ahead and a thousand other things. We move from individual ants infesting our roads with opaque objectives to orchestrated fleets, aligned to meet their common goal; green waves, road trains, speed matching and dynamic lane allocation all become much more practical once you can contact the vehicle to tell it how you would like it to behave. Of course, such a model demands a much more complete traffic management infrastructure than we have today – we need to move from sensing what has happened on the network and advising drivers by means of lights and signs, to planning what will happen on the network and providing ‘flight plans’ to individual vehicles. Such a migration promises huge increases in available network capacity and the ability to adapt to incidents much more readily to mitigate their impact.
How are we going to deal with the insurance of the vehicle, and what should the liability models look like? At the very least we should be expecting some significant changes here – it’s not reasonable to expect an individual to underwrite the behavior of a vehicle that’s acting autonomously, so OEMs will end up taking a much larger role. Its not quite as straightforward as ‘It’s the manufacturers problem’ though, because commands are still given by the owner; Remember that car that’s going to pick you up after your night out on the town? What if you asked it to come when there were five inches of snow on the ground? Even worse, what if there were five inches of snow on the ground somewhere along the route, but you’d got no way of telling from your comfy seat in the restaurant? As we stand today its easy to tie any simple liability model in knots, and we suspect there are a few lawyers going to make a fair bit of money out of the ADV transition. Personally, I’m looking forward to overhearing the argument between an ADV and it’s owner trying to convince it to come and pick them up. Didn’t that happen before sometime in 2001 with some pod bay doors?
Vehicle Drivers becoming Conductors
What about the drivers? According to UK DfT statistics there are some 38m drivers on the UK roads, and proportionately more in other countries. A fair number of these are primarily employed as drivers: truck drivers, delivery drivers, taxi drivers and a thousand other jobs. Some will still be needed (the goods will still need to be unpacked from the truck, after all) but we should expect the demand for professional’ drivers to considerably decrease. A fair number of these drivers are from the lesser-skilled end of the spectrum and will not easily pick up new work – what jobs, demanding similar levels of training, will these people be able to migrate into? Further, will we be expecting the people who previously used a car to drive between jobs to be working while in transit? Will photocopier repairmen be compelled to fill out their paperwork on the way to the next job?
As with any change, we should expect a curate’s egg situation: some good elements, some bad, and it’s incredibly difficult to forecast the net outcome…and could we see those Ph.D. theses please? While we’re on the subject, when do we expect this outcome to be? It seems to us that both ‘here’ (no autonomous vehicles in the road population) and ‘there’ (full penetration of autonomous vehicles) are easy enough to define, but the path from here to there is much more fuzzy. Our feeling is that full AD is a long-term objective with an arrival date further out than most are expecting but many of the benefits of AD will be delivered though ADAS – and so much so that the final step from ADAS to AD might not even be noted as a significant event. We already start to see ADAS in the population, fueled by increased data availability and connectivity and it’s undoubtedly bringing benefits in reduced impacts, insurance premiums and casualties alongside improved fuel efficiency and maintenance gains, but what would be a sensible implementation trajectory for full AD?
Full AD for any circumstance you come across is, as we’ve written before, a prohibitively huge challenge, but you’ll only be mildly surprised to learn that we actually already have full AD. We simply constrain the environment where its used to the point that the limitations are no longer a problem. The Docklands Light Railway in the UK and any number of airport shuttle transits are ample proof that AD can work in appropriately constrained, and instrumented, environments. If we follow this argument perhaps the correct path is to limit the applicability of AD in a ‘mixed mode’ implementation featuring both AD and conventional driving; You manually drive to the AD-instrumented motorway, flick into AD mode for the motorway segment of the trip, and go back into manual mode when you leave the motorway again; That might mean that 80 per cent of the journey is done in an AD mode, without having to deal with the 20 per cent of the journey where the complexity lives. This is a conservative, careful approach with clearly defined benefits – its also easy enough to back out of if it doesn’t work.
Finally, an open, if slightly mischievous, question. If AD offers such compelling benefits of increased capacity, improved safety and faster transit, why is it not already pervasive in other transport domains (air, rail) that, after all, offer a much more straightforward implementation environment with their rigid rule base, constrained deployment environment and more highly trained drivers who can participate in the transition?
Proposing the first pervasive deployment of AD in the most complex and multivariate transport mode has to seem a little perverse, doesn’t it?