Dead reckoning

Between burns and outside the range of useful celestial landmarks, a spacecraft tracks its own state by integrating the output of its inertial measurement unit (IMU): three accelerometers measuring linear acceleration, three gyroscopes measuring rotation rate. Integrate acceleration once → velocity. Integrate again → position. Combined with the orientation from [star trackers](mission-phases/star-trackers), the spacecraft can propagate its full state vector forward in time without any external help.

The fundamental problem is **drift**. Every accelerometer has a small bias (~10⁻⁵ m/s² on a navigation-grade IMU); over 1 day that bias integrated twice contributes ~3.7 km of position error. Over a 6-month cruise to Mars, dead-reckoning error builds up to thousands of km even with the best space-rated IMUs (which are far better than the consumer-grade ones in your phone). This is why every interplanetary mission carries [trajectory correction maneuvers](mission-phases/tcm) on its budget: the cruise dispersion has to be measured against ground tracking and the spacecraft has to physically adjust its course.

Modern best practice combines dead-reckoning with periodic Deep Space Network ranging (see [DSN](mission-phases/dsn)): the ground measures position to ~1 km accuracy via round-trip light time and Doppler, uplinks a corrected state, and the spacecraft restarts its propagation from a known-good fix. The combined system is far more accurate than either component alone. Star trackers tell the spacecraft which way it's pointing; the IMU tells it what's happening between fixes; the DSN tells it where it actually is.

Dead-reckoning matters most when the DSN isn't available — during a Mars-EDL communications blackout, during the ~22 minute one-way light delay to Jupiter and beyond, or during an attitude-control crisis. Every spacecraft autonomy system is essentially a state-vector propagator running on dead-reckoning while ground operations work on the recovery.