Sensor Category Selection
For most gimbal pointing applications, including geo-referencing, the gimbal control system requires position data, which precludes a pure IMU or AHRS solution and requires a GNSS-aided option.
In manned systems, the vehicle's onboard navigation system is a safety-critical component and interfacing with it can be onerous. In unmanned systems, the autopilot or equivalent can generally maintain vehicle control with a lower-precision navigation
system than required for accurate pointing of most payloads. And in both cases, flexing of the vehicle introduces errors since the platform's INS is typically mounted near the center of mass of the vehicle and not co-located with the gimbal. Therefore
it is best that gimballed payloads incorporate their own position and attitude sensing in most applications.
A GNSS/INS provides accurate heading under dynamic conditions, where the horizontal accelerations measured independently by the accelerometer and the GNSS can be correlated to track a high-accuracy heading output - a process known as dynamic alignment.
This is well suited to fixed wing applications, however for static or low-dynamic applications such as helicopters, rotorcraft or marine vessels a dual-antenna GNSS/INS is a preferred solution as it incorporates and GNSS Compass to provide high
accuracy heading data independent of magnetometers.
Using heading and position data we are able to georeference a point. However we need to consider the contribution of uncertainties/errors in the attitude and position estimates to the accuracy of the georeferencing solution. The error budget determines
the contribution of all sources - each sensor has an inherent margin of accuracy - that affect the acquired data quality, to check if the degree of uncertainty in the measurement solution meets the minimum required specifications.
For our ISR application we need to consider contributions from:
- GNSS/INS attitude and position uncertainties
- Timing errors
Contribution from the GNSS/INS attitude and position errors relatively easily mapped to the flat DEM.
The more challenging items to address are the two sources of misalignment errors. Misalignment error is an orthogonality error caused
by the measuring axis being misaligned. In our case of a gimballed georeferencing application this is a misalignment between the camera and the navigation system (GNSS/INS). This misalignment can be broken down into its various sources:
- GNSS/INS misalignment
- Gimbal and encoder misalignment
In order to compensate for GNSS/INS misalignment errors, the system should be calibrated to determine the fixed misalignment offset by comparing it to an independent source of truth. Gimbal and encoder misalignment are a result of the flawed assumption
that all axes are perfectly perpendicular, unfortunately this is not the case due to manufacturing tolerances. Additionally encoders are limited in resolution which leads to quantization errors.
A quick example
A plane traveling at 50m/s due North at a height of 1000m, the camera is pointed at 60° directly in front of the vehicle with a theoretical slant range of 2000m.
In a flat DEM the error budget from horizontal GNSS/INS uncertainty directly comes from the GNSS horizontal position uncertainty in an NED coordinate frame. For standard GNSS this gives an error of 2m in the North and East axis. The contribution from
the GNSS altitude uncertainty gets projected onto the DEM as a function of the gimbal tilt angle as shown in the equation below. The position error contribution can be reduced to centimeter level by using RTK.
GNSS/INS roll/yaw uncertainties produce errors perpendicular to the slant range which is parallel to the DEM. Pitch uncertainties produce errors perpendicular to which need to be projected as a function of α.
The error budget contribution from time is caused by the time delay in the measurements. Since we assume that the gimbal remains fixed the error contributed by the timing is in the direction of the velocity of the vehicle, which for this example is
due North on the NED axis.
The graph below illustrates the total error budget for our application.
Figure 4: Pitch, horizontal position errors and vertical position errors.
Figure 5: Total error budget for a flat DEM at a height of 1000m with the gimbal pointed at 60 degrees