SpaceVLA: Generalized Intelligent Autonomy for Space-Based Asset Decision-Making

My proposed work is to develop an integrated foundation-model and simulation infrastructure for autonomous operations in space-based environments, initially focused on lunar surface and Moon-base applications aligned with Artemis. The effort is two-fold, pairing (1) a pretrained Vision-Language-Action (VLA) model, which serves as the decision-making "brain" for assets such as rovers, landers, satellites, and robotic manipulators, with (2) a physics-grounded world model that provides a high-fidelity, interactive simulation environment for space-specific conditions. Unlike world models designed primarily around terrestrial environments, the space world model will represent mission-relevant factors such as lunar terrain and regolith, reduced gravity, illumination, thermal conditions, dust, sensor behavior and noise, communications, power, vehicle state, operational constraints, uncertainty, and faults. The VLA will be developed as a general pretrained model that downstream customers and stakeholders can fine-tune and post-train for their specific hardware, procedures, and objectives, then deploy within the world model to generate scenarios and synthetic training data, conduct reinforcement learning, and test and validate autonomous behaviors before flight. This workflow has produced state-of-the-art results and step-function gains in autonomy and efficiency across both autonomous driving and humanoid robotics, and preliminary research indicates it can be applied to space-based environments to greatly enhance autonomous operations. In addition to the models themselves, the work will establish reusable infrastructure, interfaces, and workflows that let NASA missions and both internal and external stakeholders adapt, evaluate, and deploy these capabilities for their own assets, creating a scalable autonomy foundation that can extend from lunar operations to Mars, orbital systems, and other deep-space environments. This effort is being pursued in collaboration with NVIDIA.

VLA model for autonomous decision-making in rovers, satellites, and landers; real-time autonomous mission operations without continuous Earth-based control; reinforcement learning-based refinement across diverse mission scenarios; establishes GSFC as the frontier center for space-based autonomy via ASTRA.

This is a mature technology (TRL 7+) — the realistic path in is usually NASA's Technology Transfer Program: licensing an existing NASA patent, or a Space Act Agreement to use NASA facilities/expertise directly. NASA also runs a startup licensing program with no upfront fee for companies formed to commercialize a specific NASA technology.

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.