SharpGeoID — Precision GNSS

SharpGeoID precision positioning

REAL GNSS

SharpGeoID fuses every GNSS signal your device can see into a single, ultra-precise position. Tap start to lock your fix.

How it works: your phone's GNSS chip already fuses GPS, GLONASS, Galileo & BeiDou. SharpGeoID reads that fused fix via the browser's Geolocation API and presents it as one sharp, confidence-scored position.

On the satellite views: web browsers don't expose per-satellite data (PRN/elevation/SNR), so those panels show a realistic multi-constellation simulation that demonstrates how SharpGeoID triangulates and converges. Your position is real; the satellite geometry is illustrative.

SharpGeoID Fix the best fused position

LOCKED

⟡ SharpGeoID Position ⟡

Latitude—

Longitude—

accuracy —precision —

Altitude

—

± Alt acc

—

Speed

—

Heading

—

Sats used

—

HDOP

—

Fix type

—

Confidence

—

Steep-climb beat grain the sub-meter reality

YOUR CASE

Grade12 %

Rider mass100 kg

Speed4.5 km/h

Heart rate160 bpm

Per-beat grain

—

Power (grav)

—

Beats / 20m pitch

—

Need 1m GPS?

At ~0.5 m/beat, beats ARE sub-meter apart — you were right. But gradient holds for ~20m pitches, so in one pitch you get ~40 beats on the SAME gradient value. Your job is to attribute beats to segments, not measure position per beat. A 2–5m map-matched fix lands every beat on the right segment unambiguously.

Do NOT compute gradient from per-beat GPS elevation — phone altimeter is ±5–10m. Use the road geometry (OSM/DEM) or a barometric altimeter; each beat inherits its segment's gradient. That's how 1m grain is satisfied without 1m GPS.

Observed precision measured, not claimed

collecting…

Reported acc

—

Observed σ

—

CEP95 (real)

—

Median jump

—

Scatter ellipse

—

HDOP surrogate

—

Reported accuracy is the number the phone claims. The other five are measured from your actual fix stream — they show the true noise floor. When CEP95 ≪ reported accuracy, the phone is being honest; when it's much larger, the phone is overconfident.

For HR-vs-gradient on a climb: once this CEP95 is in the 2–5 m range, map-match your track (next card) and you have reliable beat-to-gradient attribution. True 1 m needs RTK corrections — see the literature card below.

Road snapping · HMM map matching Newson & Krumm 2009

WORKS IN-BROWSER

raw GPS (±~10 m) HMM-snapped (on road) road graph

Raw off-road RMS

—

Snapped off-road

—

Max residual

—

Along-road σ

—

Same points, two tracks. The road graph is a physical constraint: you cannot be 6 m off the road in a garden, so every off-road residual is error the road erases. With OSM road geometry via Overpass this runs on your real recorded track — no native code, no RTK.

The 3 proven routes to ~1 m with literature

EVIDENCE-BASED

1 · HMM map matching — snap to the road graph. Along-track accuracy 2–5 m from 10 m raw GPS, because the road is a hard constraint. Newson & Krumm (2009), "Hidden Markov Map Matching Through Noise and Sparseness," ACM SIGSPATIAL. Review: Quddus, Ochieng & Noland (2007), IEEE TITS. Sufficient for HR-vs-gradient.

2 · RTK / NTRIP corrections — the only route to true 1 m and below. Raw carrier-phase + pseudorange vs a base-station network. Takasu, RTKLIB (2009). On phones: Suzuki et al., Smartphone Decimeter Challenge, ION GNSS+ 2021 — decimeter on dual-freq L1+L5 handsets. Needs native raw measurements + a dual-freq phone.

3 · GNSS + IMU + odometry fusion — Kalman tight-coupling; wheel-speed sensor on a bike is an excellent odometry channel. Groves (2013), "Principles of GNSS, Inertial, and Multisensor Integrated Navigation Systems." Bridges tree/urban dropouts.

Honorable mention: RTS Kalman smoothing on the recorded track (post-ride, zero hardware) usually lands at 2–3 m when combined with map matching.

Sky plot every satellite in view

SIMULATED

GPS GLONASS Galileo BeiDou

Signal strength dB·Hz per satellite

SIMULATED

Bars above the green line (≈30 dB·Hz) are used in the SharpGeoID fusion; weak signals are rejected.

SharpGeoID triangulation how the fix converges

SIMULATED

lines of position (per sat) SharpGeoID convergence uncertainty ellipse

Each used satellite contributes a noisy line of position. Where they all overlap is the SharpGeoID fix. More good satellites = tighter overlap = smaller uncertainty ellipse = sharper position. That's why SharpGeoID is always reliable.

SharpGeoID confidence live reliability

EXCELLENT

Fuses reported accuracy + number of satellites + geometry (HDOP) + signal quality into one 0–100 reliability score. SharpGeoID stays high because it rejects bad geometry and weak signals.

Satellites constellation detail

SIMULATED

Session

Updates

Best acc

—

Runtime

SharpGeoID

—

SharpGeoID — precision GNSS positioning. Position data is real (device GNSS via the Geolocation API). Per-satellite views are a realistic multi-constellation simulation, because browsers do not expose raw satellite measurements. Not a navigation/safety device.