Purpose This paper aims to propose an adaptive control framework for mobile manipulators to seamlessly balance autonomous task execution with human intervention during physical human–robot interaction. The objective is to coordinate a nonholonomic mobile base and a redundant manipulator in response to force-amplitude-based human guidance. Design/methodology/approach The approach integrates state-dependent dynamical systems for autonomous task encoding with a variable admittance control scheme for reactive interactions. To address the distinct kinematic constraints of the mobile platform and the end-effector, a configuration-dependent damping adjustment method and an external-force-based velocity arbitration mechanism are developed. These methods dynamically modulate the reference velocity generated by the dynamical system and the force-induced admittance velocity, enabling adaptive velocity fusion based on real-time force signals. Findings Simulation and real-robot experiments using an xArm7 manipulator mounted on a SMART mobile platform demonstrate the efficacy of the proposed framework. The results indicate that the mobile manipulator exhibits bounded and responsive behavior under the tested conditions, effectively reconfiguring to accommodate human guidance while maintaining task velocity. Originality/value This work treats the mobile manipulator as a unified redundant system rather than two separate entities. By combining dynamical-system-based trajectory generation with a novel velocity arbitration law, the proposed method realizes real-time, force-mediated coordination of the entire system. This offers a possible solution for high-degree-of-freedom mobile manipulators performing collaborative tasks in industrial environments.
PurposeThis review paper aims to provide a comprehensive review of Dubins robot path planning, which has become increasingly important in mobile robotics, autonomous vehicles and industrial applications. The authors identify key developments, analyze existing approaches and highlight the fundamental constraints, algorithms and future opportunities involving curvature-constrained, forward-only path planning. Design/methodology/approachThe review consists of three parts: basic Dubins theory, seven decades of development and future opportunities. In the first part, the authors summarize problem formulations and representative computational strategies/algorithms. In the second part, the seventy years of development are divided into four stages, with each stage summarizing developments in important topics such as point-to-point planning, obstacle avoidance and variable speed planning. In addition, the authors analyze current challenges and evaluate recent extensions in cooperative, real-time and learning-based planning frameworks. FindingsDubins path planning has evolved from simple analytical models to complex multirobot and dynamic scenarios. While exact solutions exist for basic tasks, the scalability, obstacle constraints and real-time responsiveness remain active challenges. Promising future directions include integrating machine learning for adaptive planning, and extending to Reeds–Shepp and nonholonomic systems in three-dimensional environments. Originality/valueThis review offers a structured and in-depth overview of path planning for Dubins robots, connecting core theoretical concepts with the latest developments and real-world implementations. Unlike prior surveys that roughly treat general nonholonomic planning or sampling-based methods, this paper for the first time emphasizes the curvature-constrained planning, especially under the Dubins constraints. By identifying underexplored areas such as dynamic environments, real-time applications, this review offers actionable insights and future research directions valuable to both researchers and practitioners in mobile and industrial robots.
Purpose This paper aims to present a hierarchical whole-body control (H-WBC) framework for humanoid shelf-picking in structured shelf environments. The objective is to improve motion coordination, posture regulation and safe task execution for humanoid manipulation in spatially constrained workspaces. Design/methodology/approach The proposed framework combines hierarchical whole-body kinematics, a unified task-constraint formulation and a strict-priority hierarchical quadratic programming scheme. End-effector tracking, waist posture regulation, arm-motion regularization, postural-stability constraints, joint-motion limits and shelf-related collision avoidance are integrated into a common velocity-level optimization framework. The method is evaluated in simulation and on the UBTECH Walker2 humanoid robot through single-arm and dual-arm shelf-picking tasks. Findings The proposed framework achieves accurate end-effector motion and coordinated whole-body behavior in constrained shelf environments. In simulation, it maintains low tracking errors in representative trajectory-following tasks, preserves postural stability and improves task success while reducing redundant arm motion in dual-arm shelf-picking across different shelf heights. In real-world experiments, it demonstrates practical execution of both single-arm pick-and-place and dual-arm box retrieval tasks, while maintaining stable and collision-aware whole-body motion near the shelf structure. Originality/value This study develops a reusable H-WBC framework tailored to humanoid shelf-picking. The proposed formulation unifies task objectives and physical constraints for shelf-picking within a strict-priority optimization hierarchy, and is validated in simulation and real-robot experiments on representative single-arm and dual-arm tasks in structured storage environments.
Purpose The purpose of this paper is to tackle the prediction accuracy degradation in physical contact human-robot collaborative motion due to frequent motion target changes and redundant historical data interference, by proposing a Gaussian mixture model (GMM)-Naive Bayes based motion prediction method.Design/methodology/approach The proposed method consists of a Gaussian mixture prior model and Bayesian online inference. In the prior modeling stage, a GMM is used to estimate the parameters of Gaussian components corresponding to each potential target point, thereby establishing a probabilistic mapping relationship between "target-observation." During the online inference stage, the posterior probability of the current observation data belonging to each target is calculated using the Naive Bayes method, and the target point with the maximum posterior probability is output as the prediction result. To further enhance real-time performance and prediction accuracy, a velocity-based dynamic prediction horizon adjustment strategy is introduced, which adaptively adjusts the prediction horizon according to the collaborator's real-time motion velocity. Finally, motion target prediction experiments in contact human-robot collaboration scenarios were conducted to validate the proposed method.Findings The experimental results demonstrate that the proposed prediction method exhibits strong performance in terms of changing-target adaptability, prediction accuracy and real-time responsiveness. It effectively accomplishes motion target prediction in collaborative scenarios.Originality/value A GMM-Naive Bayes hybrid prediction framework is proposed to address the issues of decreased prediction accuracy caused by target change and historical data redundancy. Furthermore, a velocity-based dynamic prediction horizon adjustment mechanism is introduced to enhance the real-time capability and accuracy of predictions in contact collaborative tasks.
Purpose This paper aims to present a novel method and tool for evaluating the motion resolution of industrial robots, defined as the minimum effective motion increment, to enable behavior-informed design and optimization of high-precision industrial robotic systems. Design/methodology/approach A high-precision laser interferometer is used to measure the joint motion resolution of a KUKA KR210 R2700 Prime robot, following an improved test procedure derived from the ISO 230 standard and an objective error-based identification method. The resulting joint maps are then used to build a prediction tool, implemented in Python and integrated with RoboDK to propagate joint-level resolution to pose- and direction-dependent end-effector Cartesian resolution estimates across the robot workspace. Findings Results show that ISO 230 tests can mask intrinsic joint resolution in the micro-step range due to backlash-driven response flattening. The proposed test procedure highlights marked joint-dependent performance, with the first three joints generally exhibiting finer micro-step resolution and more stable step responses than the wrist joints. The Cartesian predictor reproduces measured micro-displacement trends with uncertainty envelopes. Originality/value The paper provides an objective and repeatable approach that decouples intrinsic resolution from backlash, and introduces an efficient prediction tool for Cartesian resolution estimation, supporting improved system design as well as robot motion planning and compensation for precision manufacturing tasks.
Purpose To enhance the efficacy of rehabilitation training for patients with waist disorders, this study aims to develop a wire-driven waist rehabilitation training robot (WDWRTR), proposes intelligent control strategies and experimentally validates the functionalities of the WDWRTR.Design/methodology/approach The mechanical design of the WDWRTR accounts for rehabilitation movement patterns and tackles the issue of wire interference. To address the motion coupling problem in the wire traction mechanism, the kinematic model was established using two coordinate systems and the model was constructed using the Newton-Euler method. Due to the dynamic configuration of the WDWRTR, a sliding mode variable structure control method with continuous switching was proposed and its stability was analyzed. Both simulation and prototype experiments were conducted using left and right lateral bending of the waist as examples in this study.Findings The simulation outputs of wire length and tension were consistent with prototype measurements, and the actual trajectory of the scaled model closely matched the preset trajectory. Prototype experiments also validated multiple WDWRTR functions, including information communication, data acquisition and human-machine interaction. The developed prototype can perform the lateral bending movements required for waist rehabilitation training. Overall, the results demonstrate the effectiveness of the proposed control method.Originality/value This study highlights the scientific significance and reference value of the WDWRTR in flexible rehabilitation robot research by providing valuable evidence for subsequent structural optimization and control algorithm iterations.
Purpose This paper aims to address the dependence on high-definition maps, complex road networks and temporary obstacles encountered in autonomous inspection tasks in industrial parks. Design/methodology/approach A hierarchical path planning framework based on lightweight topological graphs is proposed. At the global planning layer, a directed topological road network is constructed from pre-collected road trajectories, and a global reference loop is generated using an improved Johnson-cycle method (I-JCycle) under a start-node strong-connectivity constraint. At the local planning layer, candidate trajectories are generated using Hermite curves and evaluated through a cost-based selection mechanism for obstacle avoidance and safe stopping. Findings Real-vehicle experiments conducted in an industrial park show that the proposed method can stably generate high-coverage reference loops without relying on high-definition maps. Compared with representative baseline methods, the proposed framework achieves a better balance between route coverage, loop compactness and planning efficiency at the global planning layer, while providing smoother and safer obstacle avoidance with better reference-loop consistency at the local planning layer. Research limitations/implications The method has been validated only in a limited number of representative industrial-park scenarios and assumes that the lightweight topological road network remains broadly valid during operation. Larger-scale validation under road closures and long-term environmental changes is still needed. Practical implications The proposed method can reduce reliance on high-definition maps and provide a feasible planning solution for low-speed inspection vehicles in semi-structured environments such as industrial parks and factories, while supporting both repeated loop inspection and local obstacle avoidance. Social implications The proposed method may support safer and more consistent autonomous inspection in industrial environments, thereby reducing reliance on frequent manual patrols in repetitive tasks and contributing to more efficient intelligent operation and maintenance. Originality/value This paper presents a task-oriented global-local hierarchical path planning framework for low-speed industrial inspection vehicles. The framework combines loop-oriented global planning on a lightweight directed topological graph with reference-consistent local obstacle avoidance. It reduces dependence on high-definition maps and is validated through real-vehicle experiments in an industrial park.
Purpose Achieving high-precision path tracking on uneven and complex terrain remains a significant challenge for mobile robots. The uneven changes in the ground and friction coefficients may introduce stochastic nonlinear disturbances, leading to degradation of the robot’s tracking performance, and even system instability. This paper aims to propose a robust predictive control strategy to address the performance degradation problem caused by such interference, with strong potential for applications in autonomous vehicles, robotic inspection and mobile robotics in uncertain environments. Design/methodology/approach This study models the interaction between the robot and the ground as a class of stochastic nonlinear functions. To ensure efficient implementation within embedded systems, the controller is synthesized by minimizing a regularized quadratic performance index. Furthermore, the stability of the closed-loop system is verified by Lyapunov stability analysis and linear matrix inequality (LMI) conditions. Findings Experiments demonstrate that the proposed method achieves high-precision tracking on marble, carpet and asphalt surfaces, exhibiting robust performance in environments with disturbances and uncertainties. Originality/value The control framework proposed in this study enhances the system’s robustness and computational efficiency by combining stochastic nonlinear modeling with regularization design, offering significant industrial application value.
Purpose Mobile robot localization in structured environments, such as warehouses, is frequently hindered by ghosting inherent in maps generated via 2D Laser Range Finder (LRF)-based SLAM. To address this issue, this study aims to introduce an enhanced localization framework specifically engineered to mitigate these ghosting effects. By effectively filtering out map noise, the proposed method significantly improves positioning robustness and precision, overcoming the limitations often encountered by traditional systems in such scenarios. Design/methodology/approach The proposed method introduces a novel system to construct optimized feature maps from raw SLAM output, effectively filtering ghosting noise while significantly reducing data volume. Subsequently, the Iterative Closest Point (ICP) algorithm is used to register real-time laser scans against this feature map for precise pose estimation. The framework’s efficacy is validated through comprehensive comparative evaluations in both simulated environments and real-world physical scenarios against conventional baselines. Findings Experimental results demonstrate that the proposed method significantly outperforms conventional approaches. In real-world tests, the root-mean-square error (RMSE) for translation and rotation was reduced by 0.034 m and 0.6, respectively. Furthermore, the system showed improved efficiency, with computational time decreased by nearly 8.2%. These findings confirm that the feature map-based approach effectively mitigates ghosting, enhancing both localization accuracy and operational efficiency. Originality/value This study addresses the persistent issue of map degradation in 2D SLAM without relying on complex multi-sensor fusion. By innovatively optimizing map representation through a specialized feature map generation mechanism, the method simultaneously compresses data and eliminates perceptual aliasing. This offers a robust, cost-effective solution for industrial automation in complex, structured environments.
Purpose Traditional neck-shoulder rehabilitation devices exhibit limited interaction fidelity due to rigid mechanical structures and imprecise multi-axis synchronization when interfacing with nonlinear human soft tissue. This study aims to develop a dual-motor neck-shoulder massage robot featuring kinematic optimization and a hierarchical adaptive control architecture to achieve compliant force tracking and high-precision motion coordination. Design/methodology/approach A crank-rocker mechanism, synthesized via kinematic optimization, generates biomimetic trajectories, complemented by a scissor-lift module for active depth adjustment. Within the control framework, an extended Kalman filter fuses multi-modal sensor data for real-time contact state estimation. An incremental Fuzzy PID controller accommodates the nonlinear stiffness of muscle tissue to ensure active compliance. Concurrently, a Robust Adaptive Cross-Coupling Synchronization (RACCS) algorithm regulates dual-motor coordination under variable loads. Findings Validation with 45 subjects demonstrates a 30% tracking error reduction compared to open-loop baselines on the same hardware. Compliant force control precision is maintained within a 0.2 N margin, yielding over 90% target area coverage across the cervical and shoulder regions. Stability analysis confirms the robustness of the RACCS algorithm against heterogeneous load disturbances. Originality/value This study contributes a kinematically optimized crank rocker and scissor lift mechanism together with a hierarchical adaptive control architecture for distributed dual motor systems. The integrated design manages nonlinear soft tissue impedance and offers a scalable platform for precise cervical fatigue relief in healthy subjects.
Purpose This paper aims to address the “last millimetre” problem in mobile robot manipulation, where current mobile manipulators can only align to workplaces with limited accuracy, particularly in angular alignment. The authors introduce a novel self-reflective visual servoing system that significantly improves alignment repeatability, enabling mobile manipulators to perform tasks with precision comparable to fixed robots. Design/methodology/approach The authors developed a visual servoing system called “Selfie Aligner” that uses a camera tool mounted on a robot arm and a calibration tag with integrated mirror. The system uses optical principles to achieve high-precision alignment without requiring complex camera calibration. The methodology uses transformation matrices to determine the relative position between robot and workplace, using centre-point detection of circular patterns and self-reflection for precise angular alignment. Findings Experimental results demonstrate that the Selfie Aligner improves angular alignment accuracy significantly compared to traditional methods. The system achieves a repeatability of less than 0.1 mm in position and 0.3 mRad in orientation. A practical test with markings done post alignment at 1 m from the tag showed no observable difference in the marks left after three individual calibrations. Originality/value The presented approach introduces three novel concepts: a self-reflective visual servoing method that eliminates the need for complex camera calibration, a robust alignment strategy using circular patterns that is inherently resistant to lens distortion and a simplified yet highly accurate pose estimation method using mirror-based self-reflection. This system enables mobile manipulators to achieve fixed-robot precision levels while maintaining simplicity and robustness in industrial environments.
Purpose This paper aims to solve the attitude control problem of the unmanned aerial manipulator’s (UAM) systems under dynamic coupling disturbances, thereby improving the system’s dynamic performance, disturbance rejection capability and stability in complex operating conditions. Design/methodology/approach Taking a UAV equipped with a 2-DOF manipulator as the research object, a complete dynamic model of the coupled system is established. A hierarchical antidisturbance control architecture is designed: the UAV position loop and the manipulator joints adopt nonlinear disturbance observer (NDOB)-PID control, while the attitude loop uses the proposed NDOB-ADRC strategy. The performance of the attitude-loop control strategy is systematically evaluated through multiple sets of comparative simulations, including step response, disturbance torques induced by manipulator motion and composite operating conditions with noise. Finally, the entire system is verified via integrated simulation control. Findings Experiments demonstrate that under compound disturbances, the NDOB-ADRC attitude control strategy achieves a mean absolute error below 0.0033 rad, a root-mean-square error below 0.0333 rad and a high-frequency energy ratio in the attitude error below 8.03% across all attitude channels. Compared with traditional ADRC and NDOB-PID controllers, the proposed strategy exhibits the best overall performance in all tests. Subsequently, an integrated control simulation of the UAM system was conducted under prescribed flight conditions, verifying the feasibility of the proposed hierarchical control architecture. Originality/value The innovation of this paper lies in integrating the NDOB with ADRC to propose an NDOB-ADRC strategy specifically designed for UAM attitude control and in constructing a coupled dynamic model and a hierarchical antidisturbance control architecture for this system.
Purpose The purpose of this study is to address deployment-oriented perception for single-branch deformable linear object (DLO) routing in cabinet-like industrial cells with pre-mounted arrow-shaped brackets, focusing on two practical bottlenecks: obtaining real-time, background-robust DLO masks after one-time initialization and estimating a graspable bracket pose under self-occlusion when the dominant residual ambiguity is the roll angle about the local DLO axis. Design/methodology/approach A deployment-oriented closed-loop pipeline is proposed: Grounded Segment Anything Model (GSAM, a zero-shot foundation model) is used only once to initialize the target DLO mask from a single RGB image of the scene; afterwards, a Gaussian mixture model (GMM)-based color-domain filter with lightweight post-processing (GMM-POST) performs real-time online segmentation with fixed parameters. The segmented mask is skeletonized to localize the bracket center, and a calibrated mapping from left-right visible area asymmetry is used to estimate the roll angle about the local DLO axis under top-view self-occlusion. The estimated grasp pose is integrated into sequential grasp-adjust-insert execution, forming a perception-to-action routing loop. Findings Mask IoU of GMM- POST exceeds 0.73 on simple textures DLO and 0.69 on complex textures DLO - outperforming RT-DLO and Grounded-Efficient SAM in accuracy while achieving real-time performance (34 FPS), a large speed improvement over GSAM's 0.09 FPS. Maximal roll angle-estimation error is 3.2 degrees, even when most of bracket is occluded. In the original cabinet-like setup, the end-to-end three-bracket routing task succeeds in 15 of 20 trials. Additionally, the authors conducted 30 end-to-end grasp-adjust-insert runs in an independent testbed with different robot/camera/DLO, achieving 23/30 successes, which further supports repeatability across setups and extends the evaluation to a softer-to-stiffer cable regime. Practical implications The proposed framework is readily transferable to other DLO, brackets and industrial scenarios without additional large amount of data-collection or new hardware. Originality/value This paper presents a deployment-oriented perception-and-routing pipeline for DLO with pre-mounted brackets. Its value lies in combining one-time single-image initialization with lightweight online perception, an interpretable roll-aware grasp-pose estimation module for self-occluded brackets and end-to-end validation across two physical setups with the same bracket type.
Purpose Efficient and reliable fruit handover between robotic arms is a critical bottleneck in automated harvesting systems. This paper aims to present a novel dynamic visual servoing framework that reformulates handover as an active tracking problem: a collection arm uses Position-Based Visual Servoing (PBVS) to robustly follow a fruit carried by a picking arm. This paradigm shift reduces the need for precise interarm synchronization and complex trajectory planning, offering a simpler and more adaptive solution for dynamic manipulation in unstructured environments. Design/methodology/approach The system is implemented on a heterogeneous two-arm platform, coordinated by a finite state machine. A PBVS controller is designed to maintain millimeter-level relative positioning under real-world conditions. The performance of the proposed two-arm dynamic handover system is quantitatively evaluated against a traditional single-arm serial picking approach through comparative experiments. Findings Physical experiments demonstrate that the PBVS controller achieves an average steady-state positioning error of 1.33 mm and converges within 0.85 s. Comparative results show that the two-arm system improves the picking success rate by 9.1 percentage points, reduces the total task time by 39.2% and decreases the picking arm’s movement amplitude by 39.5% compared to the single-arm baseline. Originality/value This work provides a practical, vision-driven handover solution for agricultural robotics and presents a control framework applicable to other dynamic handover scenarios. The experimental validation highlights the system’s efficiency and robustness, offering a tangible pathway toward closing the productivity gap in automated harvesting.
Purpose Autonomous mobile robots (AMRs) operating in dense and dynamically changing environments require robust navigation and reliable obstacle-avoidance capabilities. This study aims to propose an intelligent navigation framework to enable AMRs to navigate efficiently in cluttered environments.Design/methodology/approach The Cascade Neuro-Fuzzy (CN-Fuzzy) framework uses vision, LiDAR and ultrasonic sensors to detect environmental obstacles. A cascade neural network analyzes distance measurements from the sensors to calculate an optimal turning angle for the AMR, enabling it to follow a target path. Subsequently, a fuzzy logic controller generates velocity commands to ensure smooth and adaptive robot motion. The proposed system is evaluated through MATLAB simulations and real-time experimental validation.Findings The CN-Fuzzy design navigated successfully in cluttered environments. The path length error was 2.85% in unknown environments, 2.98% in indoor environments and 3.37% in complex situations. The motion time error decreased to 1.61% in unknown situations, 2.66% in indoor environments and 3.22% in complex scenarios. The proposed system achieves an average path error of 3.07% and an average motion time error of 2.50%, with corresponding root mean square error values of 2.39 cm for path length and 0.22 s for navigation time and mean square error values of 5.71 cm & sup2; and 0.047 s & sup2; for path length and motion time, respectively. These findings demonstrate the system's real-time path tracking and obstacle avoidance capabilities. Its lower error rates, enhanced robustness and smoother linear and angular velocity variations in both experiment Scenario I and Scenario II make it well-suited for precision and time-sensitive AMR navigation tasks.Originality/value This research introduces a CN-Fuzzy control framework fusing neural learning and fuzzy control for the adaptive navigation of AMRs. This hybrid approach, unlike conventional control approaches, improves situational awareness and decision-making, thus enhancing operational capabilities in densely cluttered static and dynamic environments by allowing AMRs to better interpret sensor data and respond to unexpected obstacles in real time.
Purpose This paper aims to present an automated programming system for the painting of structured industrial workpieces. The system aims to address the challenges posed by multisource uncertainties, such as unknown workpiece models or deviations in digital models during the programming of conventional painting robots. Design/methodology/approach This work combines high-precision 3D point cloud sensing devices with deep learning technology to enhance the system’s adaptability. Surface geometries are captured via line-structured light scanning, followed by targeted denoising and preprocessing. A hybrid segmentation approach, incorporating geometric feature learning and PointNet++-based instance segmentation, is used to classify the point clouds into four distinct surface types. In addition, hierarchical trajectory planning methods are developed for efficient painting. Findings Simulations and experiments demonstrate that the average deviation of paint distribution is less than 7%, with a coverage rate exceeding 99%. These results confirm the system’s effectiveness and practicality in improving paint quality under challenging environments. Originality/value This study proposes an automatic robotic painting system that significantly enhances adaptability for industrial workpieces with model uncertainties, offering a solution with real-time perception-to-action capability. The deep learning segmentation and trajectory planning methods introduced in this work can be extended to other industrial robotic applications requiring adaptive processing of unmodeled geometries.
Purpose This paper aims to address the challenges of wheel slip, instability and low precision encountered by conventional robots on unstructured rebar meshes in major infrastructure projects like nuclear power plants and wind power foundations. A novel wheel-legged hybrid tying robot with high environmental adaptability was developed. The proposed robot, optimized via a coordinated multilevel approach, successfully integrates lightweight design with robust traversability and precise operational stability on flexible rebar grids.Design/methodology/approach A hybrid mobile platform integrating a six-wheeled longitudinal self-guiding mechanism and a two-legged lateral tumbling mechanism was developed. A multilevel optimization framework was constructed: the first level used the Solid Isotropic Material with Penalization-based topology optimization to determine the macro-configuration of the frame; the second level introduced Response Surface Methodology for multi-objective fine-tuning of plate thicknesses. Finally, ADAMS dynamic simulations and physical prototype tests were conducted to verify motion stability.Findings Results indicate that the robot possesses excellent terrain adaptability, with a stable longitudinal angular velocity of 270 degrees/s and a mean velocity of 200 mm/s with minimal fluctuation. Multilevel optimization reduced the frame from 23.65 to 15.65 kg, achieving a cumulative weight reduction of 33.8% while exceeding design thresholds for strength and stiffness. Prototype experiments confirmed that the robot completes four continuous node ties within 20 s, validating the design and optimization strategy.Originality/value This study proposes a unique wheel-legged adaptive configuration and a multilevel optimization method for unstructured environments. It resolves the contradiction between lightweight design and dynamic stability, providing a reliable platform and theoretical reference for intelligent construction in major infrastructure.
Purpose This study aims to efficiently establish an accurate static stiffness model for closed-loop parallel manipulators while ensuring practical identification efficiency. Design/methodology/approach An equivalent joint stiffness identification method is proposed in this paper. This method incorporates the contact interface stiffness between internal components within each joint during the stiffness modeling. The entire system is then equivalently modeled as a hypothetical robotic system containing actuated and constrained joints, whose stiffness parameters are parameterized using high-order polynomial functions. Following this, an indirect loading experiment scheme and an identification-pose optimization strategy are proposed for stiffness parameter identification to improve identification efficiency and reduce the reliance on traditional multidimensional loading and dense multipose experiments. Findings The proposed method enables stiffness parameter identification with a reduced experimental burden while maintaining modeling accuracy. Experimental studies on a parallel manipulator at independent validation poses demonstrate that the identified stiffness model provides reliable stiffness prediction. Originality/value This study integrates internal joint contact interface stiffness into a system-level equivalent stiffness identification framework and combines unidirectional loading with pose optimization to improve identification efficiency for closed-loop parallel mechanisms.
Purpose The wrist flexibility of an industrial manipulator and flexible objects in standard assembly tasks, such as a peg-in-hole or beam-in-slot, poses challenges. Residual vibrations from high-speed motion increase assembly time and negatively affect overall productivity. This study aims to propose an innovative vision-based control strategy to manage flexibility and reduce assembly time. Design/methodology/approach The control strategy suppresses robot vibrations without altering the dynamics model or the internal controller. Its stability and bounds are determined, and its effectiveness is validated through peg-in-hole experiments. A low-frame-rate camera serves as a vision sensor for low-cost automation. Image processing and Python are used to detect vibrations in the object. Findings A pure wrist actuation control strategy is used to suppress vibration, reducing the vibration amplitude by approximately 96% within 3 s, leading to an approximately 96.38% reduction in stability time during the peg-in-hole assembly task. In addition, the strategy conserves approximately 65.14% of energy and reduces carbon emissions in repetitive industrial operations. Practical implications This method reduces vibrations and can be used in industrial robots during the assembly of objects with stepped cross sections. Suitable for peg-in-hole assembly, spot welding on stationary jigs or PCB inspection tasks. Originality/value This work examines both flexibility and stiffness using a low-cost vision camera with a pure wrist control strategy. Energy consumption is reduced by an average of approximately 65.14% due to reduced arm movement compared with other control methods, without affecting the robot’s internal controller.
Purpose Sensing constraints, strong velocity-tension (V-T) coupling, and rail-defect-induced disturbances make V-T coordinated winding control challenging for open-type abrasive-belt rail grinding robots. This paper aims to propose a tension-observer-enhanced variable-gain bi-layer active disturbance rejection control (VGBL-ADRC) framework to mitigate V-T coupling and achieve coordinated winding control under external disturbances.Design/methodology/approach The model-based tension observer reconstructs belt tension online from encoder-derived kinematics, which provides sensorless feedback and eliminates the need for dedicated tension sensors in the rail-grinding environment. Building on this, the error-dependent variable-gain bi-layer ADRC is implemented in the velocity and tension loops. Nonlinear extended state observers incorporating smooth error-driven gains, in conjunction with the corresponding variable-gain feedback laws, not only enhance the disturbance rejection capability and tracking performance but also simultaneously constrain noise amplification. Such a synergistic effect mitigates dynamic coupling between belt velocity and tension, thereby facilitating coordinated V-T winding control.Findings ADAMS-Simulink co-simulation under corrugation and localized defects shows that the belt speed fluctuation is kept within 2% and the maximum speed error is limited to 0.4%. Track experiments further demonstrate the advantage over conventional ADRC: in forward grinding, the proposed method reduces the velocity RMSE by 52.9% and the tension RMSE by 40.9%, and lowers the steady-state MAE by 71.9% in speed and 70.9% in tension. Similar improvements with consistent trends are observed in reverse grinding.Originality/value The proposed strategy provides a practical sensorless solution for disturbance-dominated, strongly coupled winding control in abrasive belt rail grinding and can be extended to other harsh industrial grinding environments.
