Efficient Adaptive Learning for Robotic Operations in Unmodeled Environments with Dynamic Uncertainties

The use of fully autonomous systems is necessary in situations where human intervention is impractical, too slow, or too tedious. For instance, a rover on Mars must explore autonomously since communication delays prohibit human teleoperation. Similarly, a free-flying robot aboard Gateway should autonomously perform routine inspections while the space station is left unattended. Achieving precise and efficient control of these systems is essential to ensure high-performance task execution while minimizing resource consumption. Controllers are designed to optimize performance based on established system designs and operating conditions. However, performance may deteriorate unexpectedly due to modelling errors, changing dynamics, or inadequate simulation or testing environments. For example, a quadrupedal robot initially ground tested on Earth might encounter unexpected terrain challenges on the Moon. Similarly, a free-flying robot may undergo wear and tear that impacts its actuation dynamics. Hence, it is crucial for these systems to autonomously adapt to varying or unmodeled conditions. The goal of this work is to develop algorithms for robotic systems which learn from past experiences to adapt to changing dynamics and environments that are difficult to simulate. Methods will be devised to learn unmodeled dynamics and physics-based models will be leveraged to accelerate learning. Learned components will be used to fine-tune control policies for robotic systems operating in complex, high-dimensional environments. Emphasis will be placed on conducting experiments with real-world hardware to validate the effectiveness of the algorithms developed through this work.

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.