Dave Marples suggests that a slow and steady approach to autonomous vehicles might actually get us there more quickly.
Article from Thinking Highways, February 2015 - also available in pdf
The celebrated US Army General Creighton Abrams famously said that “When eating an elephant, take one bite at a time”. One of the biggest elephants I’ve spotted recently in the Intelligent Transport arena is the craze for autonomous driving. My colleague Andy Graham has already indulged in a spot of big-game hunting on this subject, so I’ll refrain from re-enforcing his sentiments using different words, but the whole autonomous driving media frenzy does make me wonder where all our transport engineers went…or if they’re simply not being listened to.
Any experienced transport engineer will tell you that it would be disproportionately risky and expensive to go from our current mechanism of conducting vehicles on the road, to a completely automated approach in some single generation big bang. Not only that but we know perfectly well, from years of experience, that the challenge to achieve any big societal change is not technical, it’s social, political and economic – there’s plenty of insurance, legal and behavioral challenges to be bottomed out yet, so let’s not get too hung up on the technology for the AV, which is only one relatively small part of the whole ball of wax.
No need to rush
I’ve talked about this plenty of times before, but legacy systems pay our wages, and simply replacing them wholesale isn’t an option. That statement is true no matter whether we’re talking about computers, political systems or autonomous vehicles – to get to there from here, no matter where there may be, is often via a circuitous and suboptimal route, given that the tyres are to be changed while the truck is still rolling.
The media and our own preference for the new and shiny is as much to blame for this as anything else. We don’t really like the ‘slow-and-steady-wins-the-race’ approach as the big bang always looks more impressive. I’m as much a sucker for this as anyone else.
If you believe the Google-watchers (note, not Google themselves, who appear to be being much more sensibly cautious) then we’re on the brink of a revolution that will result in autonomous vehicles on every street. Unfortunately, as Andy already pointed out, that isn’t going to happen in the near future. Instead, we need to start thinking of the fully Autonomous Vehicle as a far point on a graph, an aspiration, rather than as an imminent deliverable. Let’s just take a single example of an issue that’s not even started to be considered yet; How do we stop kids playing chicken with vehicles they know will try very hard to stop? There’s a whole new sport right there, and there are a thousand more similar issues.
Much more likely/sensible is a model offering increasing autonomy, independence and network support for existing drivers – we already see this trend, starting with the humble SatNav we are moving towards ever-increasing technology in the car that is better informing the driver; Advanced Driver Assistance Systems (ADAS) and Connected Vehicles are the current buzzwords in the less-sexy-than-Google corners of the Intelligent Transport space, and it is far more likely that those things will be in your car in the next five years than the steering wheel will be absent.
This suits my incremental agenda well – we can try things out, learn from them and, just as importantly, have a route back to where we just came from, technologically speaking, if something doesn’t work out the way we expected it to. With my cynical commercial hat on, it also means that over time there will be a lot more spend on these systems than on moving directly to full automation, so that’s good for my career longevity.
Stepts towards autonomy
I’m not for one second saying that we shouldn’t be working towards an autonomous vehicle future – there is so much promise both in terms of safety improvements and increased utilization of infrastructure (although optimizing for both individual and group efficiency remains a hard problem) that it’s too important a trend to ignore, but it should be just that; a far point on the horizon to be aspired to, rather than a reason to delay buying a new car this year. When we eventually get to the Autonomous Vehicle do we honestly expect it to be able to conduct itself on any road in our Great Empire with equal dexterity? Or is it more likely that we’ll start off in contained environments or perhaps dedicated lanes on a couple of motorways which are “AV-enabled”, a bit like the SARTRE (SAfe Road TRains for the Environment) project demonstrated a couple of years ago?
If we believe in a long-term autonomous future, and I do, then what should we do? Well, we should start off by making a noise in order to get funding to do the fundamental research and experimentation…that bit seems to be going quite well. Next we should start building experimental systems which are analogues of the real world and experimenting in environments that are intrinsically safe – when aeronautic engineers design new planes they don’t hammer together some plywood and take it down to Heathrow Airport for a spin – they have finite element analysis, computer models, in silico simulation, scale models in wind tunnels and any number of other steps before they even pull open the doors on the workshop. Once the first prototype finally lifts off the runway the pilot can have a reasonable expectation that everything that can be done has been done to make sure the thing is safe…oh, and that runway is a fairly significant distance from anything important as well.
So why then are we drafting legislation already to allow autonomous vehicles on our roads? We should be pushing simulators, silicon test rigs and scale models, not allowing the software engineer’s “let’s see if that works” mentality to prevail. The first time an Autonomous Vehicle kills or seriously injures someone due to immature systems design the whole cause of the AV will be set back, and PR teams will earn their money. Just remember, when you tuck into that elephant, it’s pretty important to avoid getting your head stomped on after the first or second tasty morsel.
- Dave Marples
- Chief Scientist
- Send email