Spacecraft Pose Estimation for Space Situational Awareness

A fully funded PhD position for maximally 4 years.

Several thousand satellites are estimated to be orbiting Earth including operational satellites, derelict satellites, rocket bodies and other mission related objects left in orbit. Further millions of small fragmentation debris are projected but unobserved. Significant advancements in Space Situational Awareness (SSA) are required to enable active debris removal and the safe operation of future satellites. Current SSA operations make use of radio and optical telescopes with the majority of optical observations resulting in unresolved images (sub or few pixel target resolution); this limits target classification, pose estimation and orbit trajectory calculation for SSA.

Partnering with UniBern, we will coordinate high resolution resolved imagery obtained at the Zimmerwald Observatory with more conventional unresolved imagery (i.e., lightcurves). This will provide a ground-truth and basis of observations for the development of new state-of-the-art 3D reconstruction and 6 Degree of Freedom (6DoF) pose estimation algorithms.

Objectives

• 3D model available. Standard approaches from literature can be applied with minor adjustments.
• No 3D model available. Joint pose and 3D reconstruction techniques such as Neural Radiance Fields or Gaussian splatting can be applied.

• 3D model available. Conventional lightcurve inversion techniques include assuming an initial 6DoF state to generate a synthetic lightcurve for comparison to the observed lightcurve. Current methods do not employ neural networks and we foresee improvement opportunities in both the state-space search and the synthetic lightcurve generation.
• Low fidelity geometry available. We will look to further alleviate the constraints of the previous case by incorporating a 3D model optimization technique into the overall lightcurve inversion pipeline. With multiple concurrent optimization steps, this is expected to be a heavy processing pipeline. 

Requirements

• Please note that acceptance into the EPFL EDIC or EDEE schools is mandatory for this position; applicants already accepted into one of these doctoral programs is favourable.
• Masters in computer science, astronomy, engineering or a related field. 
• A strong interest in the space environment. 
• Experience programming in Python is mandatory; experience with pytorch is an asset. 
• Familiarity with rigid body 6DoF pose estimation and/or 3D reconstruction and/or trajectory estimation is an asset. 
• Self-directed problem solving skills with the initiative to tackle challenges that lack clear or documented solutions. 
• Strong verbal and written communication skills in English is mandatory; French and/or German are an asset.

Applying

References