| Περίληψη: |
Autonomous grasping with free-flying space robots encounters challenges due to dynamic coupling. This phenomenon arises from reaction-induced motions of the robotic base triggered by manipulator joint movements, leading to reduced alignment accuracy between the end-effector and the target. To address this challenge, this paper proposes a trajectory planning framework that incorporates momentum-based feedforward compensation (MFC) to enhance grasping precision in dual-arm free-floating operations. The method utilises a free-floating dynamics model to predict base motions resulting from momentum exchange between the arms and the base, and explicitly incorporates these predictions into the commanded pose of the end-effector. Furthermore, a multi-phase relative pose alignment strategy is introduced, in which attitude alignment is achieved first, followed by positional alignment with respect to a unified global reference frame. A visual feedback and local replanning mechanism is also integrated near a standoff pose, at which the robot temporarily stabilises to refine the trajectory using real-time 6-DoF relative pose feedback obtained from an eye-in-hand stereo camera. Owing to its lightweight computational requirements and avoidance of global optimisation, the proposed method enables real-time online trajectory generation and updating. Numerical simulations demonstrate that the momentum-based feedforward compensation method ensures stable convergence of the end-effector pose, even with parameter perturbations of ±10%. Compared with the particle swarm optimisation-based methods, the MFC method reduces the planning time from 248.82 s to 244.99 s to just 11.2 s and 8.0 s for the two sequential grasping phases, respectively. Moreover, the proposed method achieves substantially improvements in grasping accuracy, with end-effector position and attitude errors remaining below thresholds of 10 − 3 m and 10 − 4 rad, respectively. A total of 50 ground experiments with various initial relative poses were conducted on the Space Proximity Operation Testing platform, yielding a 96% success rate for single-arm grasping and a 94% success rate for sequential dual-arm grasping. The few unsuccessful trials were primarily attributed to system-level uncertainties, such as discrepancies in the model and the hardware, unmodelled frictional effects, residual damping from air bearings, and noise in visual sensing. In the successful trials, the relative position errors converged to sub-millimetre levels, and the relative attitude errors remained within 1°. These results demonstrate that the proposed momentum-based feedforward compensation method effectively compensates for errors induced by the robotic base, significantly improving the grasping accuracy and robustness of grasping tasks performed by space robots in free-floating conditions. • A momentum-based feedforward compensation framework is proposed for free-flying dual-arm robots. • Reaction-induced base motion is predicted and compensated in the commanded end-effector pose. • Multi-phase relative pose alignment improves dynamical feasibility and execution reliability. • A lightweight planning framework enables real-time trajectory generation. • Ground experiments conducted validate high accuracy and robustness of the proposed method. [ABSTRACT FROM AUTHOR] |