A fully Free Software/Open Source toolchain allowing practical and reliable high-precision positioning on low-cost devices.
Specifically: enable sub-decimeter positioning with a total hardware cost under USD$300 (eventually on a smartphone only) and make it widely and freely available in low-income areas.
Poor people have very limited access to precise location data. Survey-grade positioning requires expensive equipment (at least $3,000 for the most basic commercially available surveying GPS setup).
Small-scale, poor, and informal landholders have very limited ability to define, assert, and defend their property. Disputes and corruption are practically ubiquitous. While a technical solution alone cannot possibly hope to solve the societal issues around land ownership and title, enabling low-income people to access high-quality surveys is probably a necessary, if not sufficient condition for global improvement of poor landholders' rights.
We suspect that improving land surveying for property rights is likely to be one of the most impactful outcomes of access to low-cost high-accuracy positioning. However, history has shown that when given technological tools that create fundamental new capabilities, people often make unexpected use of them. A few possibilities include:
- Allow small-scale and poor landholders to define and assert the boundaries of their plots via neighborhood associations and a distributed ledger
- Improve vehicle navigation by making it less ambiguous which road or lane one is traveling on
- Enable large-scale precise surveying of natural features such as rivers to improve environmental monitoring
- Enable large-scale precise surveying of human infrastructure such as drainage and buildings to improve disaster preparedness
Of course, there is also potential for negative impact. It's easy to imagine low-cost high-precision positioning being used for, say, improved targeting of improvised weapons. However, the potential benefits appear, from an initial anlysis, to be very substantial and outweigh the potential for harm.
Current low-cost devices (smartphones in the range of $100 to $1,000) have satellite positioning capability with a rough error margin of five meters (certain devices can do a bit better under ideal conditions; the best-case scenario with a stand-alone consumer mobile device is about two meters error).
Survey-grade or geodetic satellite positioning receivers in the range of $3,000 to $50,000 can achieve positional accuracy in the range of 3-50mm. This is done by:
- Using antennas capable of receiving several frequencies of signal from positioning satellites.
- Using such antennas to estimate carrier wave phase (determining not only the position of the satellite signal but the position within the wavelength (phase) of the carrier signal)
- Using differential correction whereby measured positional errors from a satellite receive at a fixed "base station" are used to correct the positional measurements of a "rover"
- Using algorithms that account for specific types of positional error or bias in multiple positional fixes to produce statistically robust "converged" positions (Precise Point Positioning or PPP)
All of these methods are now possible to one extent or another on low-cost devices. Inexpensive external receivers (in the $250 range) can be attached to a low-end smartphone enabling accuracy in the range of a few centimeters, and even the internal GPS on many smartphones can be used with correction techniques to produce positional accuracy in the decimeter range, though this is still contingent on chipset and duty cycling or even small-scale movement of the receiver to reduce multipath errors.
Low-cost hardware is already available for high-accuracy positioning. What's primarily missing is software infrastructure.
There is no simple, easily implemented, widely available, Free and Open Source toolset available for making use of capability of currently available hardware to enable high-precision positioning.
We propose to provide such a toolset by collating existing utilities, building missing parts, and creating and documenting an overall workflow. We intend to work primarily with existing libraries such as RTKLIB and user-facing tools such as OpenDataKit.
Components needed:
- An Android library (or set of libraries) to manage incoming raw GNSS data and forward it to relevant clients
- Able to communicate with both the internal GNSS receiver and with external devices via Bluetooth, USB, and WiFI
- Able to configure and control various GNSS receivers, beginning with the ublox ZED-F9P receiver and continuing with other receivers, especially those being placed in new smartphones (essentially replacing the function of proprietary desktop-based apps like U-Center with something useable from a mobile device with sensible defaults for common positioning tasks)
- Able to receive and route NMEA messages
- Able to receive and forward RTCM messages and record RINEX data
- An early implementation target is set of OpenDataKit datatype that consists of raw GNSS data, allowing ODK users without Internet connectivity to record points, lines, and polygons that can be post-processed for high accuracy
- Able to perform real-time or post-processing correction on the phone (almost certainly using RTKLIB as the correction engine)
- An Android library providing access to external positioning data without necessitating enabling Developer Mode on the device (currently required to allow apps to access what is called "Mock Location" (which, confusingly, refers in this case to the true high-precision location) rather than the onboard Google Location service)
- An NTRIP caster to receive, organize, and broadcast base station correction information
- Provided as an installable Web service that can be deployed on any local or cloud infrastructure by users who can choose to keep it private or publicly available
- Built in such a way that makes it easy, if users choose, to freely share correction information with everyone operating in their area, making it possible for all users to have access to greater positional accuracy (essentially a peer-to-peer correction network)
- Making available Precise Point Positioning algorithms to users for placement of base stations and surveying of critical points
The simplest survey configuration is a single base station providing correction information for a single surveyor (rover).
The base station is in a known location, either because it's placed on a known monument, or because it's been left for at least a long period (up to 25 hours) to collect a "converged" position and corrected using Precise Point Positioning methods.
Correction information from the base station is either streamed to an internet site (an NTRIP caster or broadcast by a radio connection (or both).
The surveyor or "Rover" carries a GNSS receiver and makes use of the correction information from the base station. At a minimum, this can be nothing more than an Android phone provided it is capable of using raw GNSS data from the onboard receiver (most phones manufactured after 2016 with Android 7.0 or later). A very few Android devices are additionally capable of disabling the duty cycling of the GNSS receiver, improving potential accuracy. A more sophisticated and accurate rover setup would include an external GPS receiver and antenna. The Android application we propose to build will connect to the Internet to get correction information from the NTRIP caster and either provide it to the GNSS receiver (which may be able to calculate the differential corrections itself) or do the relevant calculations on the phone. The same Bluetooth link allows the phone to receive the positions (pre or post correction) from the GNSS receiver. In the event that there is no internet or radio connection to the base station, the application simply logs the raw GNSS data in a RINEX format that allows the user to apply differential correction for high accuracy after completing the survey and retrieving the logged correction information from the base station as well.
The ublox ZED-F9P receiver, a $140 chip released in 2019, toether with a $50 to $100 antenna can be connected to a low-end Android phone and achieve sub-decimeter precision using correction data (real-tie or post-processed) from a base station (which can also be a ublox ZED-F9P, antenna, and Raspberry Pi for a total cost under $300).
Additionally, as of 2016 Android allows direct access to the GNSS (Global Navigation Satellite System) data via a class representing both raw and computed satellite information. This is supported on most phones manufactured in 2016 or later and shipped with Android 7.0 or higher. The list of supported phones is growing quickly, and several are widely available in low-income countries. This means that a simple smartphone in the $100 range, without any external attachment whatsoever, can in theory be used as a rover and achieve accuracy in the decimeter range.
The GPS-enabled applications on mobile phones simply calculate their position based on the signals they receive from the satellites. They don't even record the metadata necessary to perform corrections. RTKLib can be used on Android, but it needs to be properly integrated into apps, ideally as a transparent replacement for the standard Android location API, with added functionality to allow arbitrary apps to record raw GNSS data or some subset thereof.
OpenDataKit is our first target platform. It currently saves position data as points and lines (that it calls GeoPoints and GeoTraces). These can be improved by simply increasing the accuracy of the positioning, but ideally ODK would be able to function as a fully-fledged rover application, recording raw GNSS data on the Collect platform and providing user-friendly Post-Processing Kinematic or even Precise Point Positioning correction on the back end. This requires adding a new datatype to ODK that extends the geographical capabilities and formats significantly.
| Item | Quantity | Units | Cost per unit | Cost |
|---|---|---|---|---|
| Android software engineering | 6 | months | $12,000 | $72,000 |
| Back-end engineering | 3 | months | $12,000 | $36,000 |
| Geodesic library engineering (hard math) | 3 | months | $12,000 | $36,000 |
| Project management (also codes) | 6 | months | $12,000 | $72,000 |
| Admin support | 6 | months, 50% time | $3,000 | $18,000 |
| Hardware | 1 | pile of stuff | $5,000 | $5,000 |
| TOTAL | $239,000 |