← Back to NASA Technology Projects
Completed TRL 2 (started at 2, targeting 6)
Accurate terrain modeling and its use in landmark navigation play a pivotal role in proximity operations (proxops) including Terrain Relative Navigation (TRN) and Entry, Descent, and Landing (EDL), and are currently achieved through image-based Structure-from-Motion (SfM) pipelines such as Stereophotoclinometry (SPC). Although proven effective, SPC requires significant human involvement and dependence on precise a priori knowledge, resulting in complicated mission design due to the necessity of many monofunctional reconnaissance orbits that increase overall mission cost, time, and scope. Terrestrial computer vision research has recently incepted the Radiance Field (RF); a novel SfM technique capable of producing accurate geometry, photorealistic image quality, and real-time rendering. This proposal seeks to study the radiance field as an alternative to SPC for improved topographic terrain modeling and higher fidelity landmark navigation, while simultaneously adapting current methods into GSFC proxops pipelines for in-situ, onboard, and real-time use.
Radiance fields offer an efficient means of querying terrain geometry and producing photorealistic landmark views in situ, making them a natural fit for the navigation task, and alleviating many shortcomings of the current SPC framework. For instance, the dependence on exact digital terrain models (i.e., shape models) would be eliminated, drastically reducing the amount of survey imagery and ground processes required. Similarly, the need for handcrafted, a priori landmarks for image correlation is eradicated as RF models can generate landmark views in situ. More efficient rendering procedures than current ray tracing methods combined with photorealistic quality (unlike SPC, which tends to "smooth out" prominent terrain), result in higher tracking throughput and correlation accuracy. Additionally, FPGA acceleration of Splat optimization has recently shown tremendous performance reaching hundreds of frames-per-second, demonstrating the potential for a fully onboard navigation pipeline. GSFC’s recent development of an onboard deep learning accelerator, SpaceCube LEARN (TRL 9), also allows for real-time surface geometry querying from NeRF models.
The goal of this IRAD is to equip GSFC with next-generation and autonomous landmark
navigation technologies via the radiance field. Towards this objective, this IRAD proposes four key
deliverables: (i) quantitatively analyze seven NeRF and two Splat methods against SPC and one another for terrain modeling accuracy and landmark fidelity, (ii) integrate NeRF/Splat methods into existing GSFC software pipelines, including model development with GAVIN (Goddard AI Verification and INtegration) and onboard execution with GIANT (Goddard Image Analysis and Navigation Tool), (iii) profile NeRF/Splat methods on current SpaceCube flight hardware, measuring runtime, memory usage, and power requirements, and (iv) explore additional benefits to the radiance field formulation,
including the incorporation of spacecraft, solar, and surface geometry as view synthesis priors, global
shape model creation, and SpaceCube-accelerated volumetric and rasterization rendering.
Listed on TechPort itself — the most direct way to ask about this specific project.
This is early/mid-stage (TRL 2) — the most realistic path in is NASA SBIR/STTR, which funds small businesses and research institutions to develop technology aligned with NASA's needs (equity-free, phased funding). Check whether a current SBIR/STTR solicitation topic overlaps with this project's technology area, or contact the project directly (above) to ask.
None of these are guaranteed paths for this specific project — TechPort itself doesn't have an "apply" button. Reaching out to the contact(s) above with a specific question is usually the fastest way to find out what's actually open.