GEORyder System-Level Validation of Embedded Guidance and Navigation Algorithms for Autonomous Rendezvous and Docking
Year: 2026
Type: conference-paper
Abstract: The deployment of autonomous rendezvous and docking capabilities requires Guidance and Navigation (GN) architectures capable of operating reliably under tight computational constraints and imperfect sensing conditions.Despite extensive simulation-based investigations, structured validation approaches explicitly connecting numerical modeling and embedded real-time implementation are still limited.Within the GEORyder project, a staged validation campaign has therefore been developed, integrating closed-loop Software-in-the-Loop (SIL) simulations with vision-based relevant environment experiments, enabling systematic comparison between simulated and hardware-executed performance.In particular, in the SIL configuration, an operationally representative simulation environment based on Unreal Engine reproduces the relative dynamics expected during proximity operations in Geostationary Transfer Orbits (GTO) and generates synthetic image streams consistent with the on-orbit visual conditions.These images are consequently processed by the Navigation module to successfully estimate the relative 6-DoF state of the chaser with respect to the target.The estimated state is finally fed to the Guidance module, which has the role of computing a safe approach trajectory and the associated control actions, closing the loop through numerical propagation of the dynamics.To move beyond pure simulation, the same GN software is deployed without structural modifications on an embedded development platform, where real-time operability, computational load, and numerical robustness are assessed under realistic hardware constraints.The relevant environment experimental phase concentrates on the terminal approach, starting from approximately 15 meters of relative distance from the target spacecraft and covering the critical close range segment.Scaled 3Dprinted targets, designed consistently with the optical properties of the test camera, are utilized to preserve geometric consistency with the expected operational scenario.In practice, a monocular camera mounted on a robotic manipulator observes the targets, generating the images of representative approach trajectories under varying illumination conditions.Finally, the navigation algorithm processes the acquired image streams to reconstruct the full 6-DoF relative state, allowing the quantitative evaluation of pose accuracy and sensitivity to perception uncertainties, as well as the direct comparison with SIL outcomes during the most delicate phase of docking.Overall, the SIL and close-range relevant environment activities provide a consistent view of how the GN architecture behaves when moving from simulation to hardware execution.The results suggest that the GN architecture sustains realistic sensing effects and closed-loop operation without compromising navigation performance, while the embedded characterisation quantifies the computational margin still to be recovered on the target processor, supporting its applicability to autonomous rendezvous and docking operations in GTO.
Institution: Politecnico di Milano
Language: English
Subfield: Aerospace Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7224689809