Artificial Intelligence – Simulation Based Machine Learning for Robotics Navigation

Goal: To demonstrate that machine learning is a viable approach to develop artificial intelligence for robotics system. Capability Need/Knowledge Gap: The objectives of this project are to design an intelligent agent with an artificial neural network (ANN), train the agent in simulation for resource prospecting, and deploy IA to the robot for operation demonstration. State-of-the-Art/Knowledge: Machine learning (ML) is a relatively new field of study under artificial intelligence (AI) and data science. Many big tech companies, such as Google have been investing heavily in ML research to process big data. The results are many breakthroughs, most noticeable in the housing market forecast, computer vision, and natural language processing. Key Technical Challenges: Creating high fidelity robotic & environment models in simulation, and successfully training the agent to converge at the optimal solution. Approach/Research Plan: (1) Complete robotic rover design and model testbed environment in simulation; (2) Implement artificial neural network IA in simulation and develop training methods; (3) Develop an interface to deploy IA to robotic hardware; and, (4) Demonstrate that AI can adapt to terrain with random obstacles and targets placement. We begin the ML development process with a high fidelity simulation model. This includes the robot dynamics model, the environments (known - testbed) that the mission could take place in and some anticipate objects (unknown – rocks and resources) that are placed randomly in the simulation. We will use data collected from previous reconnaissance missions to construct the environmental simulation. The training phase is where we develop the intelligent agent via reinforcement learning. An artificial brain, the Deep Neural Network (DNN), dictate an agent ‘s behavior. Finally, the agent’s brain (DNN) will be transferred to a physical robot and deploy to the environment (testbed) to carry out its mission. Next Step: To further develop this concept of using machine learning in autonomous systems, more complex operations should be investigated.

This project aligns with the Science Technology Mission Directorate’s need for game-changing system architectures to revolutionize future space missions. It addresses the agency’s need for ISRU and autonomous robotic systems. This project differs from traditional practices by utilizing artificial intelligence. There are primarily three ML paradigms: supervised learning, unsupervised learning, and reinforcement learning.

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.