Attitude and Heading Reference Systems Questions and Answers

Answer: C

Correct answer: False, they sense only the sum of the two

The statement is false. An accelerometer senses specific force, the sum of the aircraft's acceleration and the reaction to gravity, and both act identically on the proof mass - a consequence of the equivalence of gravitational and inertial mass, not of any imperfection in the instrument.

Being level does not help, since a forward acceleration tilts the sensed vector by \( \arctan(a/g) \) whatever the attitude. Nor is the limit a matter of noise, or of the two having distinguishable signatures. Separating them requires additional information - the gyros over short intervals, and an external aid over long ones.

Answer: D

Correct answer: Bias and random walk produce error growing at the same rate

This is the false statement. The two mechanisms grow quite differently: a constant bias gives attitude error growing linearly with time, while angle random walk gives error growing as the square root of time.

The other three are true. A bias integrates to \( bt \); random walk, being integrated white noise, accumulates as \( \sqrt{t} \); and a Kalman filter can indeed estimate the bias by carrying it as a state and updating it whenever the reference sensors disagree with the propagated attitude. It is precisely because the growth laws differ that the filter can separate the two.

Answer: A

Correct answer: No, movement corrupts levelling

Alignment cannot be performed while taxiing. Levelling determines initial pitch and roll by assuming the only specific force present is the reaction to gravity; any acceleration adds to that vector and tilts it, so the computed attitude is wrong. Even a slow taxi involves accelerations, turns and bumps large enough to matter.

The same objection applies to the initial bias estimates, which assume the true rates are zero. The gyros cannot compensate, since they are themselves being calibrated. Magnetic interference is a separate concern and not the reason.

Answer: B

Correct answer: Fitting a second unit

A second unit does not help identify which has failed. Two units allow a comparator to show that the outputs disagree, but the data contain nothing to say which of the two is wrong, so both must be treated as suspect.

The other three all help. A third unit makes the outlier identifiable, since it is the one disagreeing with both others; median voting among three performs that exclusion automatically; and comparison against the standby instruments provides an independent source to arbitrate between two disagreeing units. Detection needs two sources, isolation needs three.

Answer: C

Correct answer: False, gyros drift and gravity gives no azimuth reference

The statement is false, for two reasons that reinforce each other. Rate gyros measure change of heading, not absolute heading, and the integration drifts - a bias of only \( 0.5 \) degrees per hour gives a quarter of a degree in half an hour and accumulates from there. And gravity, the other reference available, is vertical, so it fixes pitch and roll but provides nothing in azimuth.

Ring laser gyros drift less but still drift, so quality does not remove the need. The only alternative to a magnetic sensor is gyro compassing on the earth's rotation, which requires navigation-grade sensors.

Answer: D

Correct answer: It avoids the need for coning and sculling compensation

This is not a reason to prefer strapdown - it is the opposite. Coning and sculling compensation exist because the sensors are fixed to a vibrating airframe and the attitude integration must be done numerically; a gimballed platform, isolated from the airframe and mechanically maintaining its orientation, does not need them.

The genuine advantages are the other three: removing gimbal bearings, torque motors and slip rings, reducing mechanical wear accordingly, and moving the coordinate transformation into software, where it is cheap once processors are fast enough.

Answer: A

Correct answer: No, an outlier pulls the average

Averaging is not better. A faulty unit's output drags the mean in proportion to how far it has strayed, so a single large error corrupts the result - with readings of \( 088.0, 088.3 \) and \( 093.5 \) degrees the mean is about \( 089.9 \), nearly two degrees from where the two healthy units agree.

The median is \( 088.3 \) and is unaffected however far the outlier lies. Averaging also conceals the fault instead of isolating it, so the failed unit continues to contribute. Computational cost is not the issue; both operations are trivial.

Answer: B

Correct answer: The magnetic variation for the route

Magnetic variation is not established during alignment. It is a property of the earth's field at a location, looked up from a stored world magnetic model and applied as a correction to convert magnetic heading to true - nothing the sensors on board can measure.

The other three are all products of alignment: initial pitch and roll from levelling on the accelerometers, initial heading from the magnetometer or gyro compassing, and initial gyro bias estimates from observing the stationary outputs. All three require the aircraft to be still, which is why alignment is done at the stand.

Answer: C

Correct answer: False, their bias is larger than inertial grade sensors

The statement is false. Micro-electromechanical sensors have larger bias and poorer bias stability than ring laser, fibre optic or other inertial-grade devices, and they are more sensitive to temperature, not less. Their appeal is size, cost and ruggedness, not accuracy.

Having no moving parts removes mechanical wear but not drift, which arises from the electronics and from the temperature dependence of the silicon structure. Stable temperature reduces the drift but does not eliminate it, which is why the filter estimates the bias continuously rather than relying on a fixed calibration.

Answer: D

Correct answer: Increasing the attitude output rate

Raising the output rate does nothing about bias. The error is \( bt \), depending only on the size of the bias and the elapsed time, so reporting the same drifting solution more often simply produces more samples of the same error. Update rate matters for following fast motion and for coning compensation, not for a constant offset.

The other three all help. Estimating the bias in the filter removes most of it and keeps tracking it as it changes; using the accelerometers as a long-term reference bounds the accumulated attitude error rather than letting it grow without limit; and selecting a sensor with lower bias attacks the quantity \( b \) at source.