To achieve centimeter-level accuracy in satellite navigation, systems cannot rely solely on standard pseudo-range "code" measurements. Code measurements track the modulated timing signals sent by satellites; while they are unambiguous, they are noisy and coarse. For high-precision applications like autonomous robotic inspection, systems must instead track the carrier phase—the actual radio wave carrying the data.
As highlighted by Zexu Du et al., the carrier wave has a tiny wavelength (typically around 19 centimeters). While a receiver can measure the fractional part of an incoming wave with millimeter precision, it has no inherent way of knowing how many whole cycles exist between the satellite and the antenna when it first locks onto the signal. This unknown number of whole cycles is the integer ambiguity. Until this ambiguity is resolved, the carrier-phase measurements provide only relative changes in distance, not absolute position.
The Mathematical Bridge: From Float to Fixed
The process of finding the true integer values begins with a statistical estimation, typically using a Kalman filter or a least-squares adjustment. This initial step ignores the requirement that the ambiguities must be integers, treating them instead as continuous decimal numbers. The result is a float solution—a vector of real numbers accompanied by a covariance matrix that describes their mutual uncertainty.
Mapping these continuous float estimates to their exact integer counterparts is mathematically challenging. As Shuguo Pan et al. demonstrate, the problem is often highly "ill-posed." Because satellites move relatively slowly across the sky, the geometric lines of sight change gradually, creating severe mathematical correlations between the estimated ambiguities of different satellites.
[Raw Carrier Phase Observations]
│
▼
┌───────────────────────┐
│ Least-Squares/Kalman │ ──► Treats ambiguities as continuous decimals
└───────────────────────┘
│
▼
[Float Solution] (e.g., N = 45231.42 ± 0.6)
│
▼
┌───────────────────────┐
│ LAMBDA Decorrelation │ ──► Rotates the error space to break correlations
└───────────────────────┘
│
▼
┌───────────────────────┐
│ Integer Grid Search │ ──► Identifies the most likely integer vector
└───────────────────────┘
│
▼
[Fixed Solution] (e.g., N = 45231) ──► Achieves centimeter accuracy
To overcome this ill-posedness, researchers widely rely on the Least-Squares Ambiguity Decorrelation Adjustment (LAMBDA) method, a framework utilized across diverse positioning architectures from precise point positioning (Lewen Zhao et al.) to relative orbit determination (Defeng Gu et al.). LAMBDA performs a clever coordinate transformation that decorrelates the float ambiguities, shrinking the search space from an elongated, highly skewed hyper-ellipsoid into a compact hyper-sphere. This allows an efficient grid search to isolate the most likely integer candidate.
Once an integer vector is found, it must pass rigorous validation tests—such as the controllable success rate tests proposed by Lei Wang 0045 et al.—to guarantee that the selected integers are statistically sound before committing to a "fixed" solution.
Why Simple Rounding Fails: Uncalibrated Phase Delays
If the underlying truth is an integer, one might wonder why we cannot simply round a float estimate of 45231.42 to 45231. The reason is that hardware and atmospheric anomalies corrupt the signal, shifting the float estimate away from its true integer value.
The primary hardware culprits are uncalibrated phase delays (UPDs) or biases. As explained by Xingxing Li et al., the internal analog and digital circuitry within both the satellite transmitters and the receiver hardware introduces fractional delays. These delays directly contaminate the carrier phase measurements, absorbing into the float solution and destroying its pure integer nature. Furthermore, emerging low Earth orbit (LEO) constellations face similar hurdles, where uncalibrated LEO hardware delays can severely degrade ambiguity resolution if left unmodeled (Kan Wang et al.).
To recover the integer property, these biases must be isolated. Two primary methodologies exist to achieve this:
- Differencing Architectures: By subtracting measurements from two different receivers tracking the same pair of satellites, receiver-specific and satellite-specific hardware biases are mathematically canceled out. Haojun Li et al. and Fuxin Yang et al. demonstrate how introducing these double-differenced constraints into absolute positioning frameworks allows users to reliably recover and fix integer ambiguities.
- State-Space Corrections: Modern Precise Point Positioning (PPP-RTK) networks use a global or regional network of reference stations to estimate these uncalibrated phase delays at the server level, broadcasting them as real-time corrections so that isolated single receivers can resolve ambiguities without differencing (Xingxing Li et al.).
The Multi-Frequency Advantage: Cascading Wavelengths
Resolving an ambiguity on a single frequency over long distances is exceptionally difficult because atmospheric errors—specifically ionospheric delays—can easily distort the signal by more than a full cycle (Yafei Ning et al.).
Modern satellite constellations mitigate this by broadcasting signals across three or four distinct frequencies. This multi-frequency deployment enables Three/Quad-Carrier Ambiguity Resolution (TCAR/QCAR), a cascading strategy that solves the integer problem step-by-step by synthesizing virtual signals with much longer wavelengths.
[Frequency 1] ──┐
├──► [Extra-Wide Lane (EWL)] ──► Wavelength: ~数 meters
[Frequency 2] ──┘ • Noise & ionosphere are small relative to wavelength.
• Resolved instantly via simple rounding.
│
▼
[Wide Lane (WL)] ──► Wavelength: ~80 cm
• Uses fixed EWL as a geometric constraint.
│
▼
[Narrow Lane (NL)] ──► Wavelength: ~19 cm
• Final high-precision solution.
By forming linear combinations of different frequencies, the authors of Triple-Frequency Code-Phase Combination Determination: A Comparison with the Hatch-Melbourne-Wübbena Combination Using BDS Signals show that one can generate an Extra-Wide Lane (EWL) signal. The EWL combination synthesizes a virtual wavelength spanning several meters. Because a multi-meter wavelength is massive compared to typical ionospheric noise, its integer ambiguity can be resolved almost instantly using simple rounding techniques.
Once the EWL ambiguity is fixed, it serves as an incredibly precise geometric constraint. As detailed by Yangyang Li et al. and Ya Fan et al., this fixed wide-lane state is used to bootstrap the resolution of the narrower Wide Lane (WL) ambiguity (~80 cm), which in turn provides the strict geometric bounds required to lock down the final, highest-precision Narrow Lane (NL) ambiguities (~19 cm). This cascading approach effectively breaks down a highly non-linear, error-sensitive search into a series of easily solvable steps, shrinking convergence times from thirty minutes down to a single epoch.
Key Foundational Work
Analysis of ill posedness in double differential ambiguity resolution of BDS (2014) — contributed 4 concept elements to this direction.
Strong Ground Motion Recorded by High-Rate GPS of the 2008 Ms 8.0 Wenchuan Earthquake, China (2013) — contributed 1 concept elements to this direction.