How Does a Spacecraft Know Which Way It Is Pointing? It Recognises the Sky

A spacecraft in free fall has no up, no horizon and no reference to anything, so knowing which way it is pointing is not automatic and has to be established continuously. That knowledge is required for pointing an antenna at Earth, aiming an instrument at a target, orienting solar panels towards the sun and firing a thruster in the intended direction. Gyroscopes measure change in orientation rather than orientation itself, so they must be started from a known position and they drift, accumulating error that grows with time. Something that measures absolute orientation directly is therefore needed, and the stars are the only reference available that is both fixed and visible from everywhere.

The sequence runs in a fraction of a second and repeats continuously:

• •A camera exposes an image of a patch of sky
• •Software locates the star images and measures their positions and brightness
• •The pattern of angles between them is computed
• •That pattern is matched against a catalogue of many thousands of stars
• •A unique match gives the direction the camera was pointing
• •Combining that with the known mounting gives the orientation of the whole craft

The accuracy achievable is extraordinary and follows from what is being measured. Stars are effectively point sources at fixed positions known to high precision, and their catalogue positions have improved enormously with dedicated astrometric missions. A camera measuring the position of a point source on a sensor can locate it to a small fraction of a pixel by fitting the brightness distribution, which multiplies the effective resolution. Averaging across many stars in one image reduces the error further. The result is orientation known to better than an arcsecond in the best instruments, which is comparable to resolving a coin at a distance of several kilometres, and that precision is what allows space telescopes to hold a target for hours.

Spacecraft carry several attitude sensors and combine them, since each has a different weakness. A sun sensor locates the brightest object in the sky cheaply and gives one direction, which is not enough on its own since the craft can still rotate about that line. An Earth sensor finds the planet's disc by its infrared emission and gives a second direction. A magnetometer measures the local magnetic field and works only close to a planet that has one. Gyroscopes measure rotation precisely over short periods and drift over long ones. Combining these in a filter that weights each by its known reliability produces an estimate better than any single instrument supplies, with the star measurement correcting the accumulated gyroscope drift.

Several conditions blind the instrument and spacecraft are designed around them. Pointing near the sun, the Earth or the moon floods the sensor, so baffles are fitted and observations are planned to avoid those directions. Rapid rotation smears the star images and prevents a match, so a separate coarse sensor must slow the craft first. Stray light from the spacecraft itself causes problems. Radiation damages the sensor over years, producing bright pixels that resemble stars and must be identified and excluded. And a craft that has lost attitude entirely must acquire a match with no starting estimate, which is a harder computational problem than tracking and is solved by dedicated algorithms.

Free fall supplies no reference, gyroscopes measure change rather than position and drift, so absolute orientation must come from outside, and the stars are the only fixed reference visible from everywhere. A camera image is reduced to a pattern of angles between stars and matched against a catalogue. Locating point sources to a fraction of a pixel and averaging across many stars gives accuracy better than an arcsecond.