This module develops the mathematical models for the observations that orbit determination filters consume. The ideal range and range-rate equations are dervied, and the real-world complications are discussed, including: light-time correction, coordinate-frame transformations (ECI ↔ ECEF, J2000/ICRF), and the various time scales (TAI, UT1, UTC, TT, TDB) needed to keep the geometry self-consistent. The measurement Jacobians for range and range-rate are derived and when the partials with respect to station coordinates, rotation parameters, or measurement biases must be included is discussed. It closes with a survey of other observable types used in practice such as: optical navigation (center-finding and landmark-based), GPS, inter-satellite ranging, radar, LIDAR/altimetry with crossovers and DDOR.

In this module the orbit determination problem becomes statistical. The Gaussian probability density function is introduced as the standard representation of state uncertainty in this course, and the method for transforming a Gaussian under linear maps is developed. The spacecraft state covariance is propagated through the dynamics using the state transition matrix. State noise compensation (SNC) is introduced, which adds a process-noise term to the covariance equation to absorb un-modeled accelerations. SNC tuning is discussed for use in real application.

This module extends the filter beyond just position and velocity. The parameters that real missions need to estimate alongside the state, such as drag and SRP scale factors, station coordinates, range and range-rate biases, and gravitational parameters are discussed. The augmented state vector is developed along with the corresponding partials. This framework is used to incorporate dynamic model compensation (DMC) as a first-order Gauss-Markov stochastic acceleration that captures whatever un-modeled forces show up, without committing to a specific physical mechanism. The corresponding covariance contribution is derived and compared against SNC as two complementary approaches to imperfect dynamics models.

The culminating assessment synthesizes the course material into an end-to-end orbit-determination exercise. You will build a Monte Carlo analysis that varies force-model fidelity, measurement geometry, process-noise levels, and SNC settings on a representative tracking scenario, then assess each modeling choice in terms of state perturbations, residual statistics, covariance consistency, and computational cost. This assessment sets up the path forward that connects this course's modeling foundation to the estimation algorithms covered in the follow-on courses of the specialization.