Position errors on a smartphone in urban areas are typically several tens of metres, compared to just a few metres in open areas. 3D mapping of the surrounding buildings can be used to aid GNSS positioning in several different ways, significantly improving accuracy.
At UCL, we combine several different techniques. Terrain height aiding constrains the 3D position solution to a 2D surface. Shadow matching, pioneered at UCL, determines position by comparing GNSS signal strength measurements with satellite visibility predictions over a grid of possible positions. Finally, 3D mapping-aided GNSS ranging enhances conventional GNSS positioning by using satellite visibility predictions to select different statistical distributions for directly-received and reflected signals
We have tested our new 3D-mapping-aided GNSS algorithms using data recorded at the City of London, Canary Wharf, the UCL campus and streets nearby, and in Peterborough. We have achieved significant accuracy improvements in both across-street and along-street directions, reducing the 2D root mean square error (RMSD) of an outdoor position from 23m to 3.7m using a consumer-grade receiver. An error reduction from 28m to 4.2m was achieved using an Android tablet equipped with smartphone GNSS technology. During the trial we also ran our algorithms in real time on a Raspberry Pi 3, providing a new position fix every second.
We are currently finalising our performance evaluation and implementing our algorithms on a smartphone. In 2018, we will extend our algorithms from a series of one-time fixes to continuous positioning, which should lead to a further improvement in accuracy. We are also talking to industry about implementing these algorithms in their products.
By using 3D building mapping and advanced algorithms, UCL has improved the outdoor accuracy of global navigation satellite systems (GNSS) in dense urban areas by a factor of five.
Authors
Paul Groves, Mounir Adjrad, Claire Ellul
