Orbit catalogs are the cornerstone of Space Situational Awareness (SSA) and, by extension, Space Traffic Management (STM). In practice, however, positional precision is often sacrificed to maximize catalog processing volume and maintain frequent updates for a large number of Resident Space Objects (RSOs). Many applications would nonetheless benefit from precise Orbit Predictions (OP), as the improved accuracy enables more fuel-efficient maneuvers for activities such as non-cooperative long-range rendezvous operations or early collision-warning systems. Catalogs themselves suffer from limited accuracy: non-cooperative maneuvering satellites can be mistakenly identified as new objects, inflating the catalog with false positives. Improving orbit accuracy directly mitigates this issue.
Ultimately, the reliability of Orbit Determination (OD) estimates is constrained by the precision of the underlying observations. For the Medium and Geostationary Earth orbit Region (MGR), the most cost-effective sensing systems are Passive Ground-Based Optical Stations (PGBOS). Their angular precision typically ranges from the arcsecond level down to about 150 milliarcseconds (mas), enabling OD and OP accurate to a few hundred meters. The recently developed CICLOPE station at ONERA reaches 50 mas, corresponding to 8–15 m at MGR altitudes. At these distances, the main modeling challenge is Solar Radiation Pressure (SRP), as its effects depend on satellite properties that are difficult to infer for non-cooperative RSOs. SRP can induce hundreds of meters of displacement over an orbital revolution, yet CICLOPE measurements demonstrate that a simple PGBOS can resolve the resulting dynamics with decameter-level precision.
This paper exploits high-precision angular measurements of non-cooperative MGR RSOs to demonstrate sub-decameter precision over observation arcs, with OP errors remaining within a few tens of meters after several days. We first justify the choice of an empirical SRP model and detail the associated assumptions. Using a standard weighted least-squares estimator and Jacobians derived from variational equations for OD, we compare our OP solutions with precise ephemerides of MGR reference satellites. We begin with synthetic overnight observations at a low acquisition rate, incorporating the characteristic noise of the CICLOPE station. Preliminary results show median OP errors of approximately 20 m after two days, with 95% of simulations remaining below 40 m. For comparison, we also evaluate performance using a simpler cannonball SRP model. Sensitivity to observation noise and to the deployment of multiple CICLOPE-like stations worldwide is also assessed. Finally, on-sky CICLOPE observations are used to validate the performance of our precise OP approach.