Speakers
Description
This work presents an innovative framework for an autonomous Rendezvous and Proximity Operations (RPO) tool, developed as a navigation subsystem within a host satellite’s GNC architecture. Targeted for In-Orbit Servicing (IOS) and Active Debris Removal (ADR) applications, the system enables reliable and precise relative navigation from long-range acquisition to final contact.
The proposed solution achieves high-performance relative position and pose estimation across both far-range and close-range regimes, while maintaining a lightweight, power-efficient, and cost-effective architecture. Unlike conventional sensing systems, which are often bulky, power-intensive and limited in operational range, the proposed framework relies on a compact suite combining onboard processing with different optical sensors.
Specifically, the system integrates visible (VIS) and thermal infrared (TIR) cameras. Visible sensors enable accurate feature recognition of the target spacecraft under favorable illumination conditions, while thermal infrared sensors ensure robust operation during eclipse phases or low-light scenarios. A key operational constraint is the exclusion of direct sunrays intrusion within the cameras’ field of view.
The acquired imagery feeds into adaptive computer vision and machine learning algorithms for target detection and pose estimation. At long distances, where the target appears as a small cluster of bright pixels, the system performs detection, identification and tracking, providing azimuth, elevation and approximate range estimates. As the relative distance decreases and the apparent size of the target increases, more advanced models are activated to extract geometric features and estimate relative pose in the 6 Degrees of Freedom (DoF).
These measurements are progressively refined from coarse to fine accuracy and fused within an Extended Kalman Filter (EKF) combining multiple estimation model outputs into consistent estimate of relative position, orientation, velocity, and angular rates, along with associated covariance information.
A key strength of the framework lies in its autonomous management of operational modes, structured into three sequential phases: Direction and Coarse Range (DCR) for initial acquisition, Position and Coarse Orientation (PCO) for intermediate refinement, and Position and Fine Orientation (PFO) for high-precision close-range operations. These modes ensure smooth and robust transitions throughout the approach trajectory without requiring continuous ground intervention.
The system is designed to support a variety of mission profiles, including the classical approach via waypoint-based “hopping” maneuvers, mid-range inspection through fly-around trajectories for target characterization, and final non-impulsive, safe approach along a validated corridor. This enables both operational safety and high-quality data acquisition for client spacecraft assessment.
A representative mission scenario and Concept of Operations (CONOPS) are presented to illustrate system performance in a realistic servicing context.
Finally, the subsystem is inherently reusable and adaptable. Its algorithms can be trained and configured on the ground using synthetic datasets and target models, allowing applications across different spacecraft, debris objects, and orbital regimes. Beyond IOS and ADR, the framework is also applicable to Space Situational Awareness (SSA), particularly for long-range detection and tracking.