Self-driving Cars of Apple Could Use Routes That Others Have Traveled


Apple has been working on some kind of autonomous car for quite some time now. The system, which is apparently an add-on to its existing vehicles, has been noted in a research paper back in November. More plans surfaced in a presentation by Ruslan Salakhutdinov, the Carnegie Mellon professor of AI, this month. According to a patent that was discovered by Autoblog, the 2015 application of Apple for an autonomous navigation system has recently been published.

The abstract of the patent describes an autonomous navigation system that generates a “characterisation” of a driving route. This route is produced when drivers navigate along the route manually. Once the said route has been manually driven for a number of times in order to satisfy a “confidence indicator,” then the car can already drive that route by itself. The characterisations can be updated as some changes in the route occur, and it can include special rules on driving that is based on “monitoring the navigation of one or more vehicles of the route.” There is also mention of some kind of remote database of routes where cars can download routes from and upload routes to, which actually sounds a bit like some sort of an extension of Apple Maps.

While the application for patent is already a couple of years old and simply represents a logic system that Apple wanted to utilise in order to inform autonomous driving, it is an interesting approach. Using driving data from many drivers can possibly help the Apple system learn routes much faster and more accurately compared to a single car with sensors alone. It is also possible that such a system will be required to work in tandem with vehicle-based sensors also.