GPS-INS Integration

Legacy context

The site’s documented heritage begins with the Position and Attitude Data System (PADS), a high-accuracy airborne direct georeferencing system that integrated GPS and inertial sensor measurements through Kalman filtering. That work extended into the GPS Inertial Data Simulator (GIDS), a bench-top tool generating real-time GPS and IMU data streams for dynamic system checkout, and the Real Time Navigator (RTN), a generalized software package enabling integration of user-defined GPS and IMU hardware without custom code.

These legacy systems share a common thread: the practical challenge of fusing GPS observations with inertial measurements to produce a stable, continuous navigation solution. The engineering focus was always on interface flexibility, real-time performance, and alignment under dynamic conditions.

That same foundation carries directly into the modern topic of GPS INS integration. Today’s systems still require the careful coupling of satellite positioning with inertial data, whether for airborne mapping, autonomous platforms, or test environments. The core questions remain unchanged: how to handle sensor data rates, timing synchronization, and filter design across diverse hardware. The heritage here provides a practical starting point for understanding how those integration principles evolved from dedicated airborne systems to broader, configurable applications.

Core Performance Benchmarks

The foundational performance figure for integrated GPS/inertial navigation comes from NASA Ames Research Center testing in 1992, where Precision (P) Code DGPS/INS positioning achieved root mean square (RMS) errors of 1 meter horizontal and 3 meters vertical [1]. These numbers represent a tightly coupled differential GPS/inertial solution and serve as a reference point for what was achievable with military-grade P-code signals before selective availability was set to zero. For engineers designing modern systems, these figures establish a baseline: unaided INS will degrade continuously when GPS is lost, while integrated solutions maintain bounded error during GPS availability [1].

The integration architecture matters as much as the raw sensor quality. Tight coupling integrates sensor data at the measurement level, whereas loose coupling uses only the outputs of INS and GPS without knowledge of internal variables [6]. The tightly coupled scheme overcomes data latency issues present in earlier architectures; reduced latency in aiding permits carrier loops to be aided and speeds up satellite acquisition [2]. This is not merely a theoretical distinction—the choice determines how quickly a receiver can reacquire signals after an outage and how well the system handles dynamic maneuvers.

Error Growth and Degradation Behavior

When GPS becomes unavailable, the INS error estimates from the integrated filter can still improve navigation performance, but the final solution degrades consistently to the level prescribed by the available aiding sensors [1]. This means an engineer must know the quality of the inertial sensors and the duration of expected GPS outages to predict whether the system will meet mission requirements. GPS itself is subject to jamming and spoofing, whereas the INS is an autonomous navigator unaffected by external influences of this sort [1]. The integration also improves estimates of acceleration, attitude, and body rates that can be used for guidance and control [1].

The effective GPS data rate is a critical design parameter. Optimization of this rate is essential for proper execution of an integrated GPS/INS in-flight algorithm, particularly because GPS attitude and heading information must be consistently available during INS outages [4]. The algorithm must determine acceptable limits of phase lag that the GPS tracking loop introduces to the flight control system during information transmission; once these calculated limits are exceeded, INS data are used to ensure continuous availability of attitude and heading information [4]. This creates a hard design constraint: the tracking loop phase lag budget must be quantified and allocated before the integration filter can be tuned.

Architecture Selection and Sensor Fusion

The decision to place GPS receivers directly inside INS boxes—an embedded GPS/INS or "EGI" architecture—provides cost, size, and weight savings [2]. When GPS data do not have to be sent outside a box for processing, a central Kalman filter can incorporate GPS measurements corrected for selective availability effects, along with other measurements such as barometric altimeter, Doppler radar, and radar altimeter [2]. This central-filter approach is characteristic of tightly coupled systems and is the dominant architecture in modern avionics.

For systems where GPS measurements are available alongside other sensors, a two-filter structure can be employed: a dynamics filter that uses GPS carrier-phase measurements to estimate velocity and other IMU errors, and a position filter that uses the velocity output of the dynamics filter and GPS pseudorange measurements [5]. This decomposition separates the high-rate velocity estimation problem from the lower-rate position update problem, which can simplify filter tuning and reduce computational load.

Performance Standards and Interface Design

The GPS Standard Positioning Service Performance Standard provides information directly applicable to designing a system that integrates GPS with an inertial sensor; see Appendix R of RTCA/DO-229 for an example [7]. This document is essential for determining whether the SPS signal-in-space can satisfy user needs and for establishing baseline GPS performance as an ICAO-standardized GNSS for both direct and augmented aerial navigation [7]. Engineers should consult this standard when defining receiver requirements and interface specifications for integrated systems.

For software architecture, modular GPS navigation software requires that GPS data interface standards be defined, with custom interface software needed to convert the output of the selected GPS receiver to a standard data interface format [8]. The major functional capabilities identified include geometric positioning—computing spacecraft position and receiver time bias by iteratively solving the navigation equations [8]. This modular approach allows the integration filter to be developed and tested independently of the specific GPS receiver hardware, which is particularly valuable when receiver selection is constrained by cost and schedule, limiting the solution space to commercially available units [3].

Practical Engineering Considerations

The integration approach for a given mission often involves adding GPS measurements to an existing onboard navigation solution as a bootstrap capability, allowing improvement of navigation capability with minimal changes to existing hardware and systems integration [3]. This incremental approach is common in upgrade programs where the inertial navigation system is already fielded and the goal is to bound error growth rather than redesign the entire navigation suite.

When evaluating whether differential GPS provides meaningful improvement over existing landing systems, studies have shown that differential GPS would provide little accuracy improvement over the current Microwave Landing System for certain applications [2]. This finding illustrates that integration decisions must be made against the specific mission profile and existing infrastructure, not against abstract performance claims.

The cumulative error behavior in integrated systems is also mission-dependent. In one indoor test with visual odometry aiding, the landing position calculated through visual odometry alone was 6.5 meters away from the measured end of the trial, with cumulative error in the final leg making evaluation difficult unless previous errors are removed [5]. This demonstrates that aiding sensors each have their own error characteristics, and the integration filter must properly weight them relative to the inertial solution and GPS measurements.

For engineers specifying integrated systems, the practical takeaway is that the 1-meter horizontal and 3-meter vertical RMS figures from the 1992 P-code DGPS/INS test represent a tightly coupled, high-accuracy reference [1]. Modern civilian systems using the Standard Positioning Service will have different performance bounds defined by the SPS Performance Standard [7]. The integration architecture—tight versus loose coupling—determines latency, reacquisition speed, and the ability to aid carrier tracking loops [2][6]. The phase lag limits of the GPS tracking loop define when INS data must take over for attitude and heading information [4]. And the overall system will degrade to the level of the best available aiding sensor when GPS is lost, so the aiding sensor suite must be designed to meet the mission's required navigation performance during expected GPS outages [1].