Sensitivity analysis of key modeling factors in angles-only relative orbit navigation for LEO applications

Angles-only relative navigation to non-cooperative targets is constrained by weak observability, placing stringent requirements on filter model fidelity. This study presents a systematic sensitivity analysis of the four key components of a batch least-squares filter architecture: analytical propagation dynamics, maneuver modeling, measurement mapping and batch configuration. Representative near-circular low-Earth-orbit scenarios are examined, including bounded and drifting relative orbits at altitudes where differential drag is significant. The analysis yields quantitative validity bounds and actionable design guidelines. Empirical differential drag modeling is shown to be essential at low altitudes, providing substantial improvement over models ignoring this perturbation. A burn-duration threshold is identified beyond which continuous maneuver modeling must replace the impulsive assumption. Exact and simplified-exact mappings from relative orbital elements to relative position maintain accuracy across the full tested range, whereas first-order rectilinear mappings are acceptable only at close range. Batch length is proven to govern fundamental observability while batch size affects noise rejection. Residual standard deviation and bias are confirmed as a practical filter health diagnostic and the elevation bias is shown to provide a geometric explanation of the estimation error. While developed for batch architectures, the propagation and measurement model recommendations apply directly to sequential filters, providing a unified reference for angles-only navigation system design.