How do you design a navigation Kalman filter for autonomous spacecraft operations?

Spacecraft navigation Kalman filter estimates position, velocity, and other states from sensor measurements. Design process: State selection - Position, velocity (6 states), Attitude/rate (6-7 states), Sensor biases, clock errors, and Consider states for unestimated parameters. Dynamics model - Orbital mechanics with perturbations, Attitude dynamics with torque models, and Process noise for unmodeled effects. Measurement model - GPS pseudorange/carrier, Star tracker quaternion, Accelerometer/gyro, and Ground tracking (range, Doppler). Implementation: Extended Kalman Filter (EKF) or Unscented KF for nonlinearity, Numerical stability (Joseph form, UD factorization), Measurement editing (outlier rejection), and Fault detection and isolation. Tuning: Process noise selection (realistic without excessive), Measurement noise from sensor specifications, and Monte Carlo validation. Onboard considerations: Computational load (embedded processor), Memory limitations, and Graceful degradation with sensor failures.