Introduction
Global Navigation Satellite Systems are extraordinary feats of engineering, providing positioning data to billions of devices worldwide. Yet despite decades of refinement, GNSS alone is not sufficient for the demanding requirements of modern autonomous systems. A single technology, no matter how mature, carries inherent failure modes - and in safety-critical applications, those failure modes are unacceptable.
The Core Limitations of Standalone GNSS
Understanding why GNSS falls short requires looking at its fundamental operating principles. A GNSS receiver determines its position by measuring the time of flight of radio signals from multiple satellites. Any factor that corrupts those signals - or prevents them from arriving - will degrade or destroy the position solution.
Signal Dropouts and Denied Environments
GNSS signals originate from satellites orbiting at altitudes of roughly 20,200 km for GPS and Galileo (Medium Earth Orbit). By the time those signals reach the ground they are extraordinarily weak - typically around -130 dBm - far below the noise floor of most electronics. Any physical obstruction between satellite and receiver breaks the link completely.
- Tunnels: A vehicle entering a road tunnel loses every satellite in view within seconds. A tunnel even one kilometre long at 100 km/h represents a 36-second total GNSS outage. Standalone GNSS has no output during this interval.
- Multi-storey car parks and underground garages: Concrete floors and reinforced ceilings block all GNSS signals. These environments are completely denied.
- Bridges and overpasses: Tall bridge structures can occlude large portions of the visible sky, pushing the number of usable satellites below the four-satellite minimum needed for a 3D position fix.
- Dense forest canopy: Tree foliage attenuates L-band signals significantly, reducing signal strength and introducing multipath from reflected ground signals.
- Urban canyons: Tall buildings in city centres create geometries where only satellites near zenith are visible. The resulting poor satellite geometry (high DOP values) inflates positioning errors even when a fix is maintained.
Multipath Interference
Even when a satellite is technically visible, the signal may arrive via reflections off buildings, vehicles, or terrain rather than - or in addition to - the direct path. These reflected multipath signals introduce ranging errors that can reach several metres in severe urban environments. Multipath cannot be fully eliminated through receiver design alone, and its effects are highly environment-dependent and time-varying.
Update Rate Limitations
Standard GNSS receivers output position at 1 Hz - one position per second. This is entirely adequate for low-speed applications such as pedestrian navigation, asset tracking, or agricultural machinery. However, autonomous vehicles operating at highway speeds (30 m/s or more), drones executing rapid manoeuvres, or robotic systems performing dynamic tasks require position and attitude updates at 50–200 Hz or higher.
Latency
Beyond the update rate, GNSS receivers introduce processing latency - typically 100–500 ms for consumer receivers, even at higher output rates. For a vehicle control loop operating at 100 Hz, this latency is a fundamental control problem. Sensor fusion with an IMU effectively eliminates this problem by propagating state estimates at the IMU rate between GNSS updates.
Accuracy Gaps in Challenging Geometry
Even in open sky, GNSS accuracy fluctuates. Satellite geometry changes throughout the day. Ionospheric and tropospheric conditions vary with weather, solar activity, and time of day. A receiver working well in the morning may exhibit significantly degraded performance under the same application conditions in the afternoon.
The Case for Redundancy in Safety-Critical Systems
Aviation, automotive, and railway standards all require that safety-critical navigation systems demonstrate fault tolerance - the ability to continue functioning correctly, or to detect and report failure, in the presence of a single-point failure. A standalone GNSS receiver represents a single point of failure. If it fails - through signal denial, spoofing, hardware failure, or constellation outage - the entire positioning function is lost.
| GNSS Failure Mode | Impact on Standalone System | Impact with Sensor Fusion |
|---|---|---|
| Tunnel/garage signal denial | Total loss of position | IMU dead-reckoning maintains position estimate |
| Multipath in urban canyon | Position jumps of metres | IMU detects inconsistency; outliers rejected |
| High-speed manoeuvre (low update rate) | Stale position, control lag | IMU propagates at 200 Hz, smooth control |
| Poor satellite geometry (high DOP) | Degraded accuracy | IMU maintains quality during poor geometry |
| Spoofing attack | Undetected false position | Fusion filter detects GNSS/IMU inconsistency |
Why Fusion Fills the Holes
Each sensor technology has its own distinct failure modes and strengths. The philosophy of sensor fusion is to combine sensors whose weaknesses are complementary - when one fails, another compensates. GNSS provides globally referenced, drift-free position over long periods. IMUs provide high-frequency, high-accuracy motion data for short periods before drift accumulates. Cameras and LiDAR work in GNSS-denied environments but require feature-rich surroundings. Together, these sensors create a navigation system that is more robust than any individual component.
Autonomous navigation - whether in a self-driving car, a delivery drone, or a precision agricultural robot - demands continuous positioning without gaps. The architecture for achieving this is sensor fusion, and GNSS is its most important, but not its only, pillar.