New Method Lets Free-Floating Robots Precisely Estimate Object Inertia in Space

A team of researchers has developed a new method to accurately estimate the inertia of objects grasped by free-floating robots. This breakthrough is particularly valuable for space robotics, where precise control in weightless or low-gravity environments is essential. The technique ensures physically realistic results, improving stability and manipulation in challenging conditions.

The group, including Akiyoshi Uchida, Antonine Richard, Kentaro Uno, Miguel Olivares-Mendez, and Kazuya Yoshida, designed the approach to work even when robots are not anchored to a surface. Their work builds on existing identification methods while incorporating momentum conservation principles for better accuracy.

The researchers used Recursive Least Squares with a novel regularisation term called Log-Determinant Divergence. This addition improves the stability and accuracy of parameter estimation, ensuring that values like the inertia tensor remain physically plausible. Testing in simulated environments, powered by the Mujoco physics engine, showed the method closely matched real-world values.

Potential applications span multiple space missions. For NASA's **Artemis programme**, floating robots assisting crewed lunar missions could rely on this technique for stable manoeuvres in low gravity. On Venus, where thick, turbulent atmospheres and extreme heat challenge robotic operations, the method could enhance stability against unpredictable conditions. It may also aid in space debris removal, allowing robots to safely approach and capture orbiting fragments without risking collisions. The team has validated the approach through simulations but plans to test it on real robotic platforms soon. Future work includes refining robot movement paths to boost estimation accuracy and integrating the system into adaptive control frameworks. Additional research will compare its performance against the Extended Kalman Filter, assess computational demands, and examine how noise and complex object shapes affect results.

This advancement provides a more reliable way to estimate object inertia in space robotics. By ensuring physical consistency, it strengthens model-based control algorithms for tasks like lunar exploration, Venus missions, and debris removal. The team's next steps involve real-world testing and further optimisation for broader application.