VBMGS: a vision-based multistep grasping strategy for a long-boom hydraulic robot system with structural flexibility

Abstract Long-boom hydraulic robots are widely used to transport heavy loads at construction sites. However, the kinematics uncertainty caused by the flexible deformation of their telescopic-boom structures poses a severe challenge to Cartesian motion accuracy, resulting in an imprecise kinematics mapping between joint space and Cartesian space. In daily life, humans naturally perform approach-to-grasp actions through experience-based approximation, vision-guided error correction, and natural compliance of their palms, even without precise arm kinematics. Inspired by human grasping behavior, we develop a hybrid-driven robotic grasping system that combines a hydraulic crane with an electric gripper and propose a vision-based multistep grasping strategy (VBMGS) augmented by a compensated incremental kinematics (CIK) model to address the above challenges. The VBMGS consists of three key steps: (1) rough approach—the robot coarsely positions itself above the target object using binocular vision; (2) precise approach—the robot refines its position and aligns it accurately with the object via monocular vision; and (3) compliant grasp—the gripper performs soft docking and secure locking to complete the final grasping. Additionally, the CIK and associated joint control methodologies are presented in detail. Four experiments, namely compliant grasping, motion sequence, secondary refinement, and autonomous grasping and handling studies, are conducted to determine key parameters and motion logic of the VBMGS and to validate the effectiveness and reliability of the developed grasping system. Results demonstrate that the VBMGS is fully compatible with the developed grasping system and successfully achieves long-distance grasping (over 10 m) with positioning accuracy maintained within the designed tolerance of 30 mm.