Currently, most mobile manipulators employ a “Stop-and-Grasp” strategy, where the base of the manipulator stops before the arm executes the grasp. However, achieving “Grasping-on-the-Move” actions—where the robot grasps a target while the base is in motion—remains a significant challenge due to the coupling of base and arm dynamics. To address this, we propose a two-phase collaborative motion planning framework. In the first phase (long-range approach), we introduce a spatially constrained visual servoing (SC-VS) method. By establishing a dynamic safety corridor based on the chassis path, this method ensures robust target tracking and obstacle avoidance for the arm during base motion. In the second phase (close-range grasping), to seize the brief grasping opportunity, we propose a Constrained-Sampling RRT-Connect (CSR-RRT-Connect) algorithm. By restricting the sampling region based on target prediction, this algorithm significantly reduces planning time. Comparative experiments demonstrate that our method achieves a 92% success rate at a base speed of 0.3 m/s, significantly outperforming the 46% success rate of baseline methods, while exhibiting superior robustness against dynamic operational disturbances and perception noise.
Sun et al. (Thu,) studied this question.