Maps, sensors and algorithms: understanding what drives an autonomous vehicle
Spanish journalist Noemí Navas called me last week to talk about the importance of cartography systems in the future of the autonomous vehicle for an article published today in El País (link in Spanish).
Our conversation focused on the companies working on digital maps, and what the commercial impact might be of the sale of Nokia Here in 2015 to a consortium of German automakers made up of Daimler, Audi and BMW, or Google’s recent divestment of its high definition satellites. Satellite maps, which already offer resolutions of less than a meter, now only require relatively routine maintenance, and are not so strategic, at least for autonomous driving.
It is important to understand that an autonomous vehicle does not depend exclusively on satellite maps for navigation. Satellite maps are a fundamental aspect of choosing a route and the ability to deal with variables such as traffic, but are not the deciding factor for the navigation itself, or decisions such as braking, accelerating or turning. These types of decisions are mainly made based on the real-time reading by the vehicle’s sensors, cameras, radars and LiDAR which provide dynamic environment control, 360º view, blind spot detection, parking assistance, lane discipline, traffic warnings at crossings, traffic light recognition, emergency braking, obstacle detection or adaptive speed control. For all these systems, the set of inputs defined by vehicle sensing is much more important than that from cartography, and are still far from fully developed, and where decisions to use sensors made by a particular supplier still have a very important impact on each company’s chances of taking the technological lead.
Similarly, data from real-time navigation of fleets of vehicles and their use to feed machine learning algorithms are also fundamental in providing an advantage to those companies able to maintain such fleets. The size of the fleet and its ability to maintain an updated system becomes fundamental, in the same way as it is in making real-time traffic maps. Communication between vehicles and between those vehicles and infrastructure, as well as between fleet and the system processing data in real time becomes key, hence the importance of elements such as 5G and data processing orientation.
The vehicle manufacturers of the future will be less bothered about design and production and more concerned with real-time data processing, which implies a change to fundamentals of the business and a transition from product orientation to offering a service that not all will be able to carry out.
(En español, aquí)
