Purpose This paper aims to investigate how redundant degrees of freedom can improve task-relevant dexterity and force capability in cooperative manipulation without disturbing commanded object motion. It compares task-specific manipulability metrics, particularly when the dominant external-load direction differs from the commanded motion direction. Design/methodology/approach A closed-loop inverse-kinematics and null-space gradient-projection framework is used to compare velocity, force and directional-force manipulability. The objectives are evaluated on a planar dual-arm hardware platform using external loads and actuator-current-based effort measurement. A complementary MuJoCo study with two 7-degrees of freedom Franka Panda manipulators extends the comparison to static holding, translational pick-and-place and six-dimensional pick-and-place from six initial object configurations. Findings Hardware experiments show that the objectives produce distinct internal configurations while preserving object motion. Velocity manipulability improves kinematic dexterity, while directional-force manipulability yields lower effort than force-only optimization when the commanded motion and external-load directions differ. The spatial simulations further show that all three objectives improve or maintain their respective measures across different initial configurations while maintaining low object-tracking errors. Originality/value The paper provides an implementation-oriented comparison of velocity, force and directional-force manipulability objectives and offers practical guidance for selecting redundancy objectives according to task and external-load direction.
Purpose This paper aims to address local path planning for autonomous mobile robots among static and dynamic obstacles, aiming to improve the robustness of the Artificial Potential Field (APF) algorithm against collisions with approaching obstacles. Design/methodology/approach A modification of the Safe APF (SAPF), called the Velocity-Scaled Safe Artificial Potential Field (VSSAPF), is proposed, in which the repulsive gain is scaled by the closing velocity, estimated online from successive distance measurements and smoothed by an exponential low-pass filter. Simulations sweep the obstacle speed from 0% to 100% of the robot’s maximum linear velocity along straight-line and circular trajectories, against potential-field baselines and the Velocity Obstacle (VO) method, followed by validation on a Husarion ROSbot 2.0 PRO. Findings Among the potential-field methods, VSSAPF attains the highest simulated success rate, 74.3% against 42.6% and 61.4% for APF with default and retuned gain, and 30.7%, 45.5% and 49.5% for SAPF, Vortex Potential Field and Adaptive SAPF. The VO reaches 100% on the straight-line trajectory, which matches its constant-velocity prediction model exactly, but only 51.7% on the circular one, where VSSAPF reaches 80.3% and is never outperformed. On hardware the mechanism is confirmed up to half the robot’s maximum speed. Beyond that every tested method collides. Originality/value The method introduces velocity-aware repulsive gain scaling requiring no prediction of the obstacle trajectory, no offline tuning and no sensors beyond distance measurements. Assuming nothing about how the obstacle moves, its performance degrades gradually rather than abruptly once the obstacle leaves a straight-line path, precisely where model-based schemes lose their advantage.
Purpose This study aims to achieve reliable sub-millimeter robotic assembly for small-scale components using vision-based control under calibration-limited industrial conditions. Design/methodology/approach A vision-based assembly strategy is proposed by integrating segmented image-based visual servoing and feature-based graph matching in an eye-in-hand configuration. The assembly state is regulated directly through two-dimensional image feature relationships without explicit 3D pose estimation. Findings Experiments on representative connector assembly tasks demonstrate sub-millimeter positioning accuracy and stable convergence under significant initial pose uncertainties using commercially available cameras, coarse calibration and reliably identifiable image features. Originality/value The proposed framework provides a practical solution for small-scale precision assembly by emphasizing feature-driven control and system-level robustness rather than high-precision sensing.
Purpose This paper aims to propose a robust, jamming-free peg-in-hole assembly control method for complex components, such as threaded screws, in unstructured environments using visuotactile sensing. Design/methodology/approach The approach introduces high-precision pose pre-alignment by performing joint estimation of the peg’s angle and offset from visuotactile depth images using a residual network. A dual-mask spatial averaging algorithm is formulated, providing a computationally efficient and low-latency solution for visuotactile force estimation. Furthermore, the search strategy integrates an active jamming recovery mechanism into an admittance-based spiral search, which leverages bidirectional transverse movements to release geometric constraints. Experiments were conducted using a 6-DOF robot arm equipped with visuotactile sensors to perform peg-in-hole tasks. Findings The proposed framework achieved a 90% assembly success rate across 60 trials, with an average hole-searching time of 11.85 s. Time-domain response experiments demonstrated that our dual-mask force estimation algorithm effectively balances real-time responsiveness and noise suppression. In the step-response experiment, the proposed dual-mask force estimation algorithm achieved a low phase lag of 53.40 ms while maintaining high signal stability (Coefficient of Variation = 0.59%). Originality/value This work contributes a complete jamming-free visuotactile assembly framework. By coupling a computationally efficient force estimation technique with a bidirectional escape mechanism, it provides a highly practical solution that enhances robotic assembly reliability.
Purpose This study aims to develop a robust vision-guided framework for instance-level recognition and grasp-based classification of metal waste in dense, multi-scale and highly reflective industrial environments, addressing the challenges of clutter, mild occlusion and dynamic operation conditions commonly encountered during the post-preprocessing sorting stage of automated recycling systems. Design/methodology/approach A lightweight two-stage What & Where to Grasp framework, termed W²G-Net, is presented. The framework adopts a cascaded pipeline integrating an optimized YOLOv11-Seg for instance-level perception and an improved Generative Residual Convolutional Neural Network (GRCNN) for planar grasp planning. An instance-mask-constrained grasp feasibility modeling method is introduced to enforce geometric and interference constraints, together with a time-compensated dynamic grasp model for conveyor belt scenarios. The approach is evaluated on metal waste segmentation data sets, a newly constructed metal waste 2D grasp data set, and the Cornell grasp data set. Findings Experimental results show that the optimized YOLOv11-Seg achieves a 17.1% inference speed improvement, while the enhanced GRCNN increases inference speed by 59.5%. The proposed two-stage framework operates at 16.2 frames per second (FPS) for single-target and 8.6 FPS for multi-target scenarios, with recognition accuracy and grasp success rates exceeding 90%. Grasp stability is significantly improved in cluttered and dynamic metal-waste environments. Originality/value This work presents a two-stage grasping framework that combines instance-level segmentation, planar grasp planning and time-compensated dynamic grasp modeling, providing an efficient, scalable and industrially applicable solution for automated metal waste sorting.
