Normal position display
For a quickly available map position, MapTools processes current GPS, network and passive Android locations. A last location no older than 30 seconds can speed up startup. This is not a statistical survey.
GPS & markers
Average and remeasure a GPS position
Multiple fresh GPS readings are filtered, robustly weighted and combined into a marker position with a transparent quality assessment.
Overview
Why use multiple readings?
A single GPS fix is a snapshot. Satellite geometry, obstruction and reflected signals can shift it. A series reveals the spread and reduces the influence of isolated outliers.
Average position
The measurement starts with an empty series and accepts only new fixes from Android's GPS provider. Old, duplicate, invalid or insufficiently accurate fixes are rejected.
Remeasure position
An existing marker can be measured again in further sessions. The new session, the robust combined result and the change in estimated accuracy remain clearly separated.
In practice
How a measurement works
The measurement completes only when minimum duration, minimum usable fix count and stability criteria have all been met.
- Choose the locationRemain at the point, preferably with open sky and away from strongly reflecting surfaces.
- Choose a profileShort, Standard or Long sets the minimum duration, required fix count and maximum waiting time.
- Remain stillMapTools collects fresh GPS fixes, checks accuracy and movement, and displays progress.
- Review the resultUsed fixes, reported accuracy, P68, P95, center stability and quality expose the underlying data.
- Save the markerAn initial measurement creates a marker. After remeasurement, the user decides whether to apply the combined result.
| Profile | Minimum duration | Minimum used fixes | Maximum duration |
|---|---|---|---|
| Short | 15 s | 10 | 30 s |
| Standard | 60 s | 30 | 90 s |
| Long | 180 s | 90 | 300 s |
Technical details
How MapTools calculates the position
The method is not a simple mean of latitude and longitude. It works in a local metric frame and combines accuracy weights with robust statistics.
1. Input validation
Invalid coordinates or accuracy values, fixes from before measurement start, duplicate timestamps and fixes with reported horizontal accuracy above 25 m are rejected.
2. Stationary check
Movement is detected when at least two of the latest three speeds exceed 1.2 m/s. With at least ten fixes, the centers of the first and last five are also compared. Limit: max(6 m; 1.5 × a_med).
3. Convert to metres
WGS84 coordinates are converted to Earth-fixed ECEF coordinates and then to a local east-north plane (ENU). Distances and weights can therefore be calculated correctly in metres.
4. Robust initial center
The separate medians of east and north form the initial estimate. The radial distance rᵢ from every fix to that center is then calculated.
5. Detect outliers
MapTools calculates MAD = median(|rᵢ − median(r)|). Outlier limit: max(3 m; median(r) + 3 × 1.4826 × MAD).
6. Weight by accuracy
For reported radius aᵢ, wᵢ = 1 / max(aᵢ; 1 m)². The ratio between the strongest and weakest base weight is capped at 4:1.
7. Huber estimate
The center is updated five times. Inside the limit the robust factor is 1; outside it becomes c / r. The limit is c = max(1.5 m; 1.5 × 1.4826 × MAD).
8. Return to WGS84
The weighted ENU center is converted back through ECEF to latitude and longitude. Available altitudes use their median; quality values assess only the horizontal position.
Result
Understanding P68, P95 and estimated accuracy
These values answer different questions and must not be treated as interchangeable.
| Display | Calculation | Meaning |
|---|---|---|
| Median reported accuracy | Median of the horizontal accuracy radii reported by Android | The provider's own estimate. Android generally defines this as a radius with about 68% probability for the individual fix. |
| P68 spread | 68th percentile of distances from the robust center | About 68% of the observed fix distances are at or below this value. |
| P95 spread | 95th percentile of distances from the robust center | About 95% of the observed distances are at or below this value. P95 = 4.2 m means roughly 95% of evaluated fixes were no more than about 4.2 m from the center. |
| Center stability | Distance between robust centers of the first and final third | A series is stable at no more than max(3 m; 0.75 × a_med) and P95 ≤ 25 m. |
| Estimated accuracy radius | max(P95_combined; median(a_session)) | MapTools uses the less favorable value so an unusually tight cloud cannot override device-reported uncertainty. |
Percentile calculation
Radial distances are sorted. With n values, the position is p × (n − 1), using p = 0.68 for P68 and p = 0.95 for P95. MapTools linearly interpolates between neighboring values.
Important interpretation
P95 describes observed spread, not a guaranteed 95% confidence radius around the true position. A tight point cloud can be shifted as a whole. Absolute accuracy is not proven without an independent reference.
Assessment
Quality levels and accuracy comparison
The quality level is based on the estimated radius. Multiple independent time windows improve the strength of the assessment.
| Quality | Estimated radius |
|---|---|
| High | ≤ 5 m |
| Good | > 5 m / ≤ 10 m |
| Fair | > 10 m / ≤ 20 m |
| Poor | > 20 m |
One session
For the combined overall assessment, MapTools awards the highest level only after two independent time windows. A single session is rated Good at best.
Improved or degraded?
A change within max(0.1 m; 0.01 × R_previous) is unchanged. Larger differences are reported by amount and percentage.
Avoid false precision
Many decimal places in coordinates do not imply matching measurement accuracy. Radius, spread, stability, environment and independent windows matter.
Multiple series
Time-separated remeasurement
Separate series can capture different satellite constellations and reception conditions, revealing effects hidden in one session.
90-minute windows
A session starting less than 90 minutes after the current window began belongs to that window. A later session opens a new one. A marker can retain up to 32 sessions.
Two-stage combination
Sessions are first robustly combined within each window, then window centers are combined. Outlier rejection also applies at three or more windows. A session center uses at least the larger of P68 and median device accuracy.
Professional reference
Garmin recommends four to eight samples at least 90 minutes apart for careful waypoint averaging. MapTools adopts the idea of varying satellite constellations but uses its own method and is not Garmin-certified.
| Action | What is stored | Marker position |
|---|---|---|
| Save and apply position | The new session is added to history. | The marker receives the robust combined result of all sessions. |
| Save without applying position | The session and combined result remain stored. | The visible marker position remains unchanged. |
| Discard measurement | The new session is not stored. | Position and previous history remain unchanged. |
No latest-session-only shortcut
Applying the combined result uses every stored session. Alternatively, the new session can be kept for documentation only or discarded completely.
Context
Limits, privacy and practical guidance
Position averaging improves statistical assessment but does not replace suitable reception conditions or independent checks for critical work.
Suitable conditions
- Prefer open sky.
- Keep away from facades, rock walls, dense canopy and vehicle structures.
- Remain at the point during a session.
- Remeasure important markers later.
Technical limits
- Multipath reception can shift an entire series.
- Android accuracy is a device estimate.
- Altitude is usually less accurate and is not part of the horizontal quality level.
- Legally relevant surveying may require geodetic equipment and prescribed procedures.
Data processing
Raw fixes are processed in memory during calculation. Only summarized session data and marker history are stored permanently. Background measurement uses a visible foreground-service status notification.
Sources
Technical sources and references
Sources are listed as plain text. They provide context for terms and methods; the specific combination and parameters are the ArxSilex MapTools implementation. Status: July 2026.
- Garmin Customer Support: “Waypoint Averaging Information for Outdoor Handhelds”.
- Garmin Support: “Waypoint Averaging on the Montana 700 Series”.
- Android Developers: “Location API reference”,
Location.getAccuracy(). - Esri: “ArcGIS Field Maps, Configure the map”, “GPS averaging”.
- Esri: “ArcGIS Field Maps, Capture”, “GPS averaging”.
- NIST/SEMATECH: “e-Handbook of Statistical Methods”, Measures of Scale, Detection of Outliers and Percentiles.
- Peter J. Huber: “Robust Estimation of a Location Parameter”, The Annals of Mathematical Statistics, volume 35, issue 1, 1964, pages 73–101.
- National Geospatial-Intelligence Agency: “Department of Defense World Geodetic System 1984, Its Definition and Relationships with Local Geodetic Systems”, NGA.STND.0036.
Manual
Related manual pages
Position averaging can be combined with marker work, coordinate checks and target navigation.