Title: Toward Autonomous Relative Navigation and Characterization of Unknown Resident Space Objects

Date: Thursday, June 25th, 2026

Time: 10:30AM to 12:30PM ET

Location: Montgomery Knight Rm 317 or via Teams (Meeting ID: 258 609 310 252 58, Passcode: n6jV6mE6)

Juan-Diego Florez

Robotics Ph.D. Student

Woodruff School of Mechanical Engineering

Georgia Institute of Technology

Committee:

Dr. Panagiotis Tsiotras (advisor) -- Daniel Guggenheim School of Aerospace Engineering

Dr. Danfei Xu -- School of Interactive Computing

Dr. Frank Dellaert -- School of Interactive Computing

Dr. John Christian -- Daniel Guggenheim School of Aerospace Engineering

Dr. Mehregan Dor -- SpaceX

Abstract:

Safe proximity operations near defunct, tumbling, or debris resident space objects (RSOs) demand onboard relative navigation without cooperative markers, target models, or IMU-based odometry. This research develops a physics-informed monocular visual simultaneous localization and mapping (SLAM) framework that embeds the coupled orbital and rigid-body equations of motion as factor-graph constraints, so that a single passive camera, aided by an attitude reference, recovers both the inspector’s relative trajectory and the target’s angular velocity, inertia tensor, and center of mass (COM) location. The objective is to characterize and extend the observability of these quantities across the full inspection regime, from well-resolved close range to partially resolved long range where the target subtends only a few tens of pixels, for passively tumbling, non-cooperative targets. The methodology combines analytic and empirical observability studies, ablations on photo-realistically rendered inspection sequences, and a trust-gated dynamics propagator that quantifies when a dynamics-based estimate is safe to consume. The work culminates in a decentralized cooperative formulation in which multiple inspectors share dynamics-consistent estimates, transfer custody across observation gaps, and establish a common target frame across viewpoints. The significance is a model-free, infrastructure-independent navigation capability for the unstructured proximity operations that future servicing, active debris removal, and space-domain-awareness missions require.