Why Autonomous Vehicles Can’t Rely on GPS Alone

Imagine an autonomous vehicle entering a dense urban environment, surrounded by high-rise buildings and reflective glass facades. Suddenly, the vehicle drifts slightly out of its lane, and the system issues a warning — “Positioning signal lost.”

At the entrance to an underground parking garage, positioning is lost instantly, and the vehicle can no longer determine whether it remains on the intended path.

If GPS is active, why does autonomous driving still lose its way?

This is because traditional GPS (or more generally GNSS) alone cannot meet the stringent requirements of autonomous driving for accurate, continuous, and reliable positioning.

How Does GPS Work?

GPS (more precisely, GNSS, including systems such as BeiDou and Galileo) determines position by measuring signal propagation time from at least four satellites.

In open environments, it can achieve meter-level to sub-meter-level accuracy.

However, real-world conditions quickly expose the limitations of GNSS:

· Urban obstructions → sharply reduced satellite visibility

· Signal reflection (multipath effects) → pseudorange measurement errors

· Tunnels / elevated roads → complete signal loss

· Dynamic motion (sharp turns, acceleration, deceleration) → GNSS alone cannot reliably estimate attitude or instantaneous velocity

As a result, the vehicle may misinterpret its position — appearing in an adjacent lane or even inside nearby buildings.

Many intelligent vehicles still rely on simplified sensor fusion approaches combining GNSS, wheel speed sensors, and low-grade IMUs. These systems may be sufficient for highway driving but often fail under the following conditions:

·   Complex urban intersections → positioning jitter exceeds 1 m, making lane-level turning decisions unreliable

·   Static conditions → no heading output (single-antenna GNSS cannot determine heading at rest)

·   GNSS outages exceeding 10 s → rapid position drift, forcing system degradation or autonomous driving disengagement

This is comparable to navigating while blindfolded — with only occasional glimpses of direction, and relying on estimation the rest of the time.

For L3+ autonomous driving, this level of risk is unacceptable.

To address this industry challenge, Archimedes Innovation has developed the PBOX tightly-coupled high-precision GNSS/INS integrated navigation receiver, delivering three key architectural-level improvements:

1. Tightly-Coupled Fusion Algorithm

Instead of loosely coupling GNSS and IMU outputs, raw measurements — including pseudorange, carrier phase, and IMU data — are tightly integrated within a unified filter.

High-precision positioning can be maintained even with as few as two visible satellites.

2. Multi-Frequency, Multi-Constellation with Dual-Antenna Heading

Supports GPS L1/L2/L5 and BeiDou B1I/B1C/B2a signals, improving robustness to ionospheric errors.

A dual-antenna baseline enables 0.1° heading accuracy, even in static conditions.

3. Automotive-Grade High-Stability MEMS IMU with Adaptive Temperature Compensation

An industrial-grade IMU with a proprietary temperature drift compensation model ensures stable attitude performance across –40°C to 85°C.

During GNSS outages, dead reckoning error remains within 0.1 m over 30 s.

The Archimedes Innovation PBOX tightly-coupled navigation receiver is not a simple GNSS + INS combination, but a tightly integrated GNSS × INS system — where 1 + 1 > 2 in practice.

The performance of the PBOX in urban testing in Beijing is summarized as follows:

Choosing the Archimedes Innovation PBOX tightly-coupled navigation receiver enables centimeter-level positioning awareness for vehicles in any environment.