Self-Driving Cars – Autonomy Pioneer Reveals ‘Dirty Little Secret’
‘The fact is, one of the dirty little secrets [of self-driving cars] is that the software today is pretty city- or country-specific.’
That’s the diagnosis of Karl Iagnemma, one of the pioneers of the technology, and the brains behind NuTonomy.
Self-Driving Cars
He’s not, however, claiming the specificity is a particular barrier or flaw in the evolution of autonomous vehicles – merely the next hurdle to be overcome. Talking to Forbes, Iagnemma said ‘As an industry we’ve really progressed through a few different phases. The first phase, to be totally honest, was to just get these cars to drive around the block – convince ourselves that the technology could work as a proof of concept.
‘We’re beyond that stage. We’re in what I would call a pre-validation stage. Validation means we’ve convinced ourselves that the software works well enough to put on the road without a driver. We’re not quite there yet as an industry, but we’re getting pretty close. That’s why you’ve got Google putting a lot of cars on the road; that’s why you have nuTonomy building up its fleet and logging a lot of kilometres.
The Next Stage for Self-Driving Cars
‘Then the next stage, the one that’s really interesting, is scaling the technology from one city to many cities. If you develop a technology that only works in a single city, where it’s kind of optimized for a specific city, that’s not really that exciting.’
He has a point – while automated vehicles only work in specific cities, for instance, the idea of robo-freight trucks remains science fiction: you’d have to unload and re-load at every city from your departure point to your destination(!). Similarly, autonomy becomes silo-based if it only works within one city, about as ubiquitous as a subway system. Multi-city applicability is the key to widespread future use.
Iagnemma sees the logic in having developed single-city usage first though – for all it needs to be opened out to usher in the future of urban transport. ‘One of the obvious reasons is driving in different cities is a distinct experience. Some of the rules of the road, the appearance of the signage, the road markings are different. So when you move to a new city, you’ve got to adapt your software to drive in that location. I think we sometimes think that once one company gets the technology right, then overnight we’ll have autonomous cars on the road in every major city worldwide. But in fact that’s not the case. It’s going to be a city-by-city, country-by-country rollout of this technology, which is going to take some time. It’s not just the rules of the road. There are cultural differences as well. In Singapore, drivers generally obey the rules, but the attitude around pedestrians is actually quite different. It’s culturally different. People drive safely, but it’s not the same deference shown to pedestrians.
Self-Diving Cars Software Development
“Depending on how you develop your software, adapting your code to go from one market to the next, it could be a fairly painless process, or it could be quite challenging and could require overhauling the code.
“Nobody really wants to hear that because with all the money they are spending on this technology, they want to believe they have a global solution at their fingertips. That may not be the case.’
One of the key technological races of the next few years then will be for software and apps that can synthesize and adapt self-driving vehicle behaviour to the ‘rules’ – both written and unwritten, both legal and cultural – of a number of cities, to speed the city-by-city spread of the technology through adaptable city profiling.