Online Inertia Parameter Estimation Enables Manipulation of Unknown Objects in Space Applications

Determining an object’s inertia is fundamental to precise manipulation, a challenge significantly amplified when robots operate in the weightlessness of space. Akiyoshi Uchida, Antonine Richard, and Kentaro Uno, from the Space Robotics Lab at Tohoku University, alongside Miguel Olivares-Mendez and Kazuya Yoshida, now present a method for accurately estimating the inertia of grasped objects during manipulation, even with free-floating robotic systems. Their work builds upon existing identification techniques, incorporating principles of momentum conservation to enable application on robots unanchored in space, and demonstrates accurate parameter estimation through detailed simulations. This achievement represents a crucial step towards more sophisticated and reliable robotic operations for both terrestrial and future space missions, allowing robots to interact with objects dynamically and predictably in zero gravity.

This research addresses limitations in traditional methods, particularly when dealing with non-cooperative targets during on-orbit servicing and active debris removal. The team’s approach utilizes Recursive Least Squares with a novel regularization term, Log-Determinant Divergence, to improve the accuracy and stability of parameter estimates. The method incorporates momentum conservation principles, specifically designed for free-flying space robots, representing a significant improvement over force-based approaches.

It also ensures estimated parameters are physically realistic, such as maintaining a positive definite inertia tensor. Future work will focus on optimizing robot trajectories to maximize estimation accuracy and integrating the framework into an adaptive control scheme. Validation has been performed through simulations using the Mujoco physics engine, and the team plans to test the method on a real robotic platform. Further research will involve comparing performance with the Extended Kalman Filter and investigating the computational complexity, sensitivity to noise, and performance with complex object shapes.

Inertia Estimation for Free-Floating Robots

Scientists have developed a novel method for estimating the inertia of unknown objects grasped by robotic manipulators, specifically for free-floating space robotics. This addresses a critical need for accurate dynamics-aware manipulation in orbital environments, where robot movements affect overall attitude. The team extended an existing online identification technique, incorporating momentum conservation for use with robots not fixed in place. Numerical simulations accurately model the interaction between a robotic arm and a target object with unknown inertia, enabling real-time parameter estimation during manipulation.

The system accounts for the coupling between manipulator motion and the robot’s base, simulating the unique dynamics of a free-floating system. The core of the method involves a recursive least squares algorithm with log-determinant divergence regularization, ensuring physical consistency and robustness. Experiments demonstrate accurate identification of inertia parameters in simulated scenarios, closely matching ground-truth values, highlighting the method’s potential for on-orbit servicing missions.

Robotic Inertia Estimation for Free-Floating Systems

Scientists have achieved accurate estimation of inertia parameters for unknown objects during robotic manipulation, a crucial capability for dynamics-aware control in space robotics. The research team refined an online identification method, extending its application to robots with free-floating bases, common in orbital scenarios. This advancement addresses a key problem where precise knowledge of an object’s inertia is essential for stable grasping and manipulation. The method leverages momentum conservation and incorporates Recursive Least Squares with log-determinant divergence regularization. Experiments through numerical simulations demonstrate the method’s ability to accurately identify inertia parameters, closely matching ground-truth values, even with unknown target objects.

This accurate identification is vital for controlling the robot arm while accounting for the resulting motion of the robot’s base in space. Measurements confirm the method’s effectiveness in scenarios relevant to on-orbit servicing and space debris capture. This delivers a robust solution for estimating inertia, ensuring physical consistency and improving model-based control algorithms, paving the way for more autonomous and reliable space robotics missions.

Robotic Inertia Estimation With Free Bases

This research presents an advancement in robotic manipulation, addressing the challenge of identifying the inertia properties of unknown objects during physical interaction. Scientists adapted an existing method for estimating inertia, extending its capabilities for robots with free-floating bases, such as those used in space applications. The team incorporated principles of momentum conservation, resulting in a more accurate and stable estimation process compared to traditional force-based methods. Results demonstrate accurate identification of target object inertia in simulated ground and orbital environments, even with errors in the robot’s own inertia parameters.

This improvement stems from the method’s ability to maintain physical consistency and reduce susceptibility to external disturbances, a critical factor for reliable operation in space. The researchers plan to test the method on a physical air-floating robot platform and compare its performance against established techniques like the Extended Kalman Filter. Future work will focus on optimising manipulator movements to enhance estimation accuracy and integrating the framework into an adaptive control scheme for real-time updates during on-orbit manipulation, with potential benefits for missions like active debris removal.