The University of California team refined an existing approach which previously needed the type of computers that have their own room.
They were trying to overcome the existing limitations by which ordinary GPS which works solely by referencing satellites is accurate to around 10 meters, while adding in reference to transmitters on the ground reduces it to around a meter.
That’s more than enough to figure out where you (or something you are tracking) are, but isn’t sufficient for some uses that require precision. A key example is autonomous vehicles where being more precise could be the difference between the car knowing it’s in the correct lane and knowing it’s positioned most efficiently for an upcoming turn. More accurate GPS could also be used for automated agriculture on large sites.
The solution the researchers worked on was already known in principle: in simple terms, it means cross-referencing the GPS data with information from an inertial measurement unit that houses equipment such as accelerometers and gyrometers. The big difficulty so far is that combining these data sources in real time to keep up with the movement has required serious computing power.
The researchers came up with a more efficient algorithm that means the technology could be handled by a device fitted in a car or, in theory at least, by a smartphone.