Track manipulation, locomotion and embodied systems.
Search papers, labs, and topics across Lattice.
This paper addresses the challenge of modeling highly nonlinear, configuration-dependent sensitivity in bio-inspired rigid-soft coupled dexterous fingers, which hinders traditional discrete-time Jacobian-based control. The authors introduce Jacobian Flow Matching (JFM), a Conditional Flow Matching framework that learns a resolution-consistent Jacobian field by treating actuation-to-motion transitions as a dynamical flow. Experiments show JFM significantly improves single-step prediction accuracy (over 53% RMSE reduction) and enhances long-horizon trajectory fidelity under sparse sampling, demonstrating its effectiveness for offline multi-step trajectory optimization in such complex systems.
A learned flow-based Jacobian field enables consistent, high-fidelity motion prediction for bio-inspired robotic fingers across varying temporal resolutions, outperforming discrete Jacobian learning by over 53% RMSE.
Bio-inspired tendon-driven rigid-soft coupled dexterous fingers exhibit strong nonlinearity and configuration-dependent sensitivity, making accurate modeling challenging. In discrete-time control, Jacobian-based kinematic algorithms typically rely on point-wise local linear approximations, which makes their performance sensitive to sensor sampling frequency and controller update frequency. To address this issue, we propose Jacobian Flow Matching (JFM), a structured learning framework based on Conditional Flow Matching (CFM), to learn a resolution-consistent Jacobian field that models actuation-to-motion transitions as a dynamical flow. The proposed framework supports both single-step prediction and continuous rollout via ODE integration, enabling consistent inference across temporal resolutions. Experiments on a tendon-driven rigid-soft finger show that the proposed method suppresses outlier errors and improves single-step prediction accuracy, reducing the global average RMSE by over 53% compared with a baseline discrete Jacobian learning approach. For long-horizon prediction, trajectories recovered via ODE integration achieve higher fidelity under sparse sampling (Stride = 8), reducing the RMSE median by 14.43% and the error variance by 24.87%. These results demonstrate that the learned flow-based Jacobian field provides an effective local model for offline multi-step trajectory optimization in rigid-soft coupled nonlinear systems.