SmartNav: Centimeter-Accurate GPS Inside the World's Densest City Streets

A smartphone dropping your pinpoint to the wrong sidewalk on a narrow downtown street is not a bug in the device — it is a fundamental limitation of how global navigation works. In tall "urban canyons," satellite signals bounce off glass towers, reflect off buildings, and arrive at the receiver from distorted paths, so the reported position can jump tens of metres. A system built at the Norwegian University of Science and Technology, called SmartNav, closes that gap by stitching together three complementary techniques into sub-ten-centimetre accuracy.

Why GPS breaks down in cities

Global navigation satellite systems were designed for open skies. On a rooftop or a highway the line of sight between receiver and satellites is clean, and standard position algorithms hold to a few metres. Between skyscrapers, the geometry collapses: the number of visible satellites drops, and the signals that do arrive are often reflections rather than direct beams. Because a reflected signal has travelled a longer path, the receiver interprets the extra travel time as extra distance and plots the user at the wrong point. The problem is acute for applications where a few metres matters — parking an autonomous vehicle, locating a delivery van in a garage, or guiding a robot on a crowded sidewalk.

The three-part fix

The approach relies on carrier-phase positioning, which reads the phase of the microwave carrier wave itself rather than just the timing code on the signal. This is what gives professional systems their precision, but it normally requires nearby reference stations and careful processing. SmartNav layers on PPP-RTK corrections — Precise Point Positioning combined with Real-Time Kinematics — to deliver the high-precision corrections over a wide area in real time from space-based and ground infrastructure, without needing a local base station.

The third ingredient is a 3D model of the city. By knowing the shape, height and location of every building along the route, the system can predict which satellite signals will be blocked or reflected, and down-weight or reject the distorted ones. The 3D building model acts as a kind of shadow map for the radio signals: it tells the receiver which measurements are trustworthy and which are echoes from a skyscraper. Combined, the three techniques convert the urban canyon from an adversary into a known, correctable environment.

What it enables

For autonomous vehicles, reliable positioning in cities has been the slow lane of the whole roadmap. Cameras and lidar handle what is right in front of the car, but they need a trusted global reference for the wider map and for fail-safe operation when the visual scene becomes ambiguous. SmartNav provides that reference at the centimetre scale, using hardware already affordable enough for mass deployment rather than purpose-built survey equipment. The result is a practical path to navigation-grade positioning on the densest streets where it matters most.

The broader picture

SmartNav shows a wider trend in navigation: the frontier is no longer raw satellite geometry, but the intelligent fusion of signals, models and correction services. As the building inventory of cities is digitised into high-resolution 3D maps, those maps become a live input to positioning rather than just a display layer. The same principle — correct the signal using the environment that distorts it — is spreading to indoor navigation, delivery logistics, and the positioning backbone of future self-driving systems.

Knowledge takeaway: urban "signal reflections and blocked satellites" degrade standard GPS to tens of metres; carrier-phase positioning reads the microwave carrier wave itself for far finer precision; PPP-RTK delivers real-time corrections without local base stations; 3D city models let the receiver discard reflected signals; together SmartNav achieves sub-10-cm accuracy with consumer-grade receivers.