HOW GOOD IS GPS?

Implementations of Differential GPS Positioning

Clearly the use of two GPS receivers, simultaneously tracking the same satellites, is an effective means of overcoming the effect of spatially correlated biases. There are essentially two ways in which measurements from two receivers are used to account for biases, and hence improve accuracy:

Differencing measurements between receivers leads to an observable that is free of biases (or at least substantially reduced if the receivers are not too far apart). This is the GPS surveying mode of differential positioning using carrier phase data.
Each set of measurements at a receiver are independently used to derive a position which is in error by more or less the same amount. This is the DGPS procedure implemented for precise navigation applications using pseudo-range data.

Relative positioning accuracies are mostly dependent on the differential technique used:

Differential GPS navigation, relative positions derived by differencing two absolute position determinations, or application of range "corrections" to measurements. Typically accurate at the few metre level, though several new procedures deliver sub-metre accuracy.

GPS surveying, where baseline components are derived from the simultaneous processing of data collected over an extended period of time (static mode), or under certain constraints (kinematic mode). Accuracy expressed as a ratio to baseline length, typically a few ppm, with a small constant term of a centimetre or less.

Several distinct categories of differential GPS positioning performance can be identified, each a function of data type (pseudo-range or carrier phase) and receiver mode (static or kinematic).

Differential Navigation Based on Real-Time Point Positioning

The distinguishing characteristic of navigation is the urgency with which positioning information is required in order to ensure safe passage from port to port. (The situation with regards to "land navigation" is slightly different from that of marine or air navigation, but the notion of keeping track of where the user is going, or has gone, is still important.) Relative positioning information is derived from two separate navigation "fixes" based on the processing of pseudo-range observations. One of these is at a fixed station of known position (the "base station"), while the other is at an unknown location (the "remote station", or if it is moving, the so-called "mobile station"). Two implementations are possible:

(1) Differential positioning can be accomplished by the continuous transmission of the coordinate solution from the "base station" to the "remote station" as illustrated in Figure 1. This block shift technique is the easiest to implement (although it does have certain limitations):

Base receiver at known point -- WGS84 or local datum coordinates.
Compare known position and instantaneously computed position.
Generate correction X, Y, Z.
Transmit correction to remote receiver for immediate correction of "raw" field coordinates (or saved to file at base for later correction of field receiver coordinates).

It is important that both the remote and base receivers use the same satellite constellation to generate their point solutions, otherwise severe errors can result, possibly worse than those of the (uncorrected) point positioning. As simple as this may sound, it can be difficult to implement in practice!

Figure 1. Principle of DGPS using the block shift method.

(2) A popular real-time DGPS strategy is the method of range corrections. Rather than making corrections to the coordinates, the ranges before computation of the receiver position are corrected (Figure 2). This is achieved by a process similar in many respects to that of block shift method:

Base station at known point --> WGS84 or local datum coordinates.
Using known position compute "true" range.
Generate corrections to all pseudo-range data by comparing "true" to "observed" range.
Transmit correction data to remote receiver for correction of ranges before solution carried out.

The technique is far more flexible because the correction is made to the pseudo-ranges and hence the remote GPS receiver can use any combination of corrected ranges to obtain a solution, and not just the satellite set used at the base station.

Figure 2. Principle of DGPS using the range corrections method.

The relative accuracies expected in such schemes are of the order of 1-10m, though enhancements have been introduced that appear to deliver submetre accuracy.

Baseline Surveying via the Combined Processing of Data from Two Phase Tracking Receivers

The distinguishing characteristics of the survey mode of positioning are:

In one scenario the data is collected simultaneously over an extended observation period, at two or more sites, whose positions do not change with time (static baseline).
An alternative scenario involves data collected in kinematic mode, at two receivers, where the remote receiver if moving and the base receiver is stationary.
The carrier phase data is used because of its very high measurement precision.
The data processing techniques involve the comparatively sophisticated modelling of the systematic errors (or biases) in the GPS system.
The vertical error is greater than the horizontal components by a factor of 2 or 3.

In tests carried out by the U.S. Federal Geodetic Control Sub-Commission (FGCS) on GPS surveying systems (section 4.4), the internal consistency (or repeatability) for static GPS baseline results were characterised by a one-sigma base positional uncertainty in each component that ranged from 3 to 10mm. Added to the base error in each component was a one-sigma line-length dependent uncertainty of 1-2 ppm. Possible sources for the base error included: phase centre variation, effects of antenna multipath, errors in centring and measuring height of antenna phase centre. The line-length dependent error was caused mostly by errors in the orbital ephemeris data and atmospheric effects. During periods of peak ionospheric activity, leading to degraded single frequency baseline results, it is likely that the line-length error budget may have to be increased slightly.

Much higher accuracies are possible when "GPS geodesy" procedures are applied. Typically relative accuracies of the order of 0.1 to 0.01ppm are obtained.

Kinematic phase-based GPS positioning is a relatively recent technique that can deliver accuracies perhaps only a few factors worse than static baseline techniques. However, there are a number of constraints:

The interstation distances must be comparatively short (generally less than 20km).
The ambiguous carrier phase data must be converted to precise "carrier-range" by the determination of the integer ambiguities before the GPS survey can commence.
In general, the most sophisticated hardware (dual-frequency receivers) is used.
The dynamics of the moving receiver must not be too great to cause significant periods of loss-of-lock on the satellite signals.

Kinematic phase-based positioning techniques are discussed in section 5.5.5.