Human–robot interaction is an important research content in the robot field. In order to achieve the compliance and cooperation of human–robot interaction process, variable parameter interaction control and contact collision response of human-redundant robot are studied. Firstly, the kinematics model of redundant robot is established, and an inverse kinematics algorithm based on the augmented Jacobian matrix is proposed with the optimization objective of avoiding joint limit. Then, the admittance control model of human–robot interaction is established, and variable admittance control based on force and velocity is proposed. The contact intention of human is judged by the vector product relationship between external force and velocity. The control parameters are adjusted to better ensure the friendliness of human–robot cooperation. When the external force is applied to robot, based on human–robot interaction control law, robot motion is controlled, and the robot can move in the direction of the force, so the goal of human–robot compliance interaction is realized. When the external force is suddenly withdrawn, by adding the stiffness coefficient, the robot moves along the direction of the force and then can return to the position where the previous force is applied, achieving the safety response to robot's contact collision.
Backdrivability can improve the safety and comfort of rehabilitation training robots, thereby facilitating active patient participation. However, existing high-backdrivability systems often result in tradeoffs, including a reduced transmission ratio or lower response bandwidth. To address these issues, a combination of a serial-parallel hybrid structure and a precision cable drive were used in this study. However, a serial-parallel hybrid structure increases the kinematic operation cost. An equivalent KP screw system method was introduced to simplify the hybrid motion chain and reduce the computational cost in the control process. The developed robot has a working space that fully covers the normal range of motion of the human upper limb and provides high motion dexterity, covering the areas required for most rehabilitation training movements. Despite its excellent reverse driving capability (10 cNm), the robot maintained a relatively high transmission ratio (1:24) and good control bandwidth (50 Hz). The rehabilitation training system design is based on the patient’s active intent, and experimental results indicate that the robot achieves high trajectory tracking accuracy (error: ±0.5°).
Quadrupedal robots demonstrate impressive mobility in unstructured environments, yet achieving robust locomotion under limited sensing remains a significant challenge. Recent progress in deep reinforcement learning suggests that proprioceptive feedback alone can support adaptive gait generation without external perception. Building on this insight, we propose a hierarchical multi-encoder actor framework for proprioception-only quadrupedal locomotion. The architecture consists of three layers: a sensor-encoder layer that independently learns velocity, contact, and predictive proprioceptive representations; a fusion layer that integrates these latent features into a compact unified space optimized via PPO; and an action layer that produces motor commands from both the fused representations and current observations. This hierarchical design facilitates modular representation learning and implicit terrain inference, enabling stable and adaptive locomotion on unseen and challenging terrains—without relying on cameras or elevation maps.
Electrical power systems are critical for socioeconomic development, and marker balls are essential for preventing collisions with low-flying aircraft. Traditionally, the installation of such markers is performed manually by technical staff, exposing them to risks such as falls, electric shocks, and physical injuries. Drones, especially when integrated with modular robots, can efficiently reach installation points along transmission lines, overcoming the limitations of manual operations. Existing drone-based solutions rely on manual control, which requires highly skilled pilots and provides limited precision, and, to date, no studies have investigated autonomous flights for this task. To address these challenges, this paper proposes an autonomous drone-robot control system for marker ball installation. The proposed algorithm combines depth image processing based on filtering and line fitting with RGB color based processing, enabling automatic landing of the robot on the power cable. More precisely, an onboard computer processes the captured images and generates setpoints sent to the flight controller, responsible for controlling the drone’s movement. Experiments conducted in a controlled environment, with an average installation time of 43 s, demonstrated the effectiveness of the proposed system. The results show that autonomous control reduces the need for pilot skill and enables safe and faster installation of marker balls.
The intensive production of potatoes ( Solanum tuberosum ) has negative impacts on human health, the environment, and the economy, mainly due to the excessive and inefficient use of agrochemicals. The implementation of mobile robots for soil inspection represents a promising technological alter-native for optimizing the management of these inputs. However, challenges remain in the modeling, development, and automatic control of this type of agricultural robot. In the field of control, strategies associated with Active Disturbance Rejection, such as Generalized Proportional Integral (GPI) control, offer high robustness and precision in trajectory tracking. However, their performance depends on correct parameter tuning. This paper proposes the multi-objective optimization of a GPI controller implemented in a Solanum tuberosum crop inspection robot. Initially, an optimization problem was formulated to simultaneously minimize tracking error and control effort, using the bio-inspired Multi-Objective Particle Swarm Optimization (MOPSO) algorithm. Subsequently, to contrast the quality of the Pareto front obtained, the evolutionary Non-dominated Genetic Algorithm II (NSGA-II) was implemented, generating an expanded set of 26 non-dominated solutions. The comparative results showed that NSGA-II presents a better distribution of solutions and greater diversity of the Pareto front, while MOPSO offers faster convergence. Together, both approaches allow for more efficient tuning of the GPI controller, improving trajectory tracking by 23 % compared to traditional methods and suggesting the applicability of hybrid optimization strategies in the automation of agricultural robots.
The intensive production of potatoes ( Solanum tuberosum ) has negative impacts on human health, the environment, and the economy, mainly due to the excessive and inefficient use of agrochemicals. The implementation of mobile robots for soil inspection represents a promising technological alter-native for optimizing the management of these inputs. However, challenges remain in the modeling, development, and automatic control of this type of agricultural robot. In the field of control, strategies associated with Active Disturbance Rejection, such as Generalized Proportional Integral (GPI) control, offer high robustness and precision in trajectory tracking. However, their performance depends on correct parameter tuning. This paper proposes the multi-objective optimization of a GPI controller implemented in a Solanum tuberosum crop inspection robot. Initially, an optimization problem was formulated to simultaneously minimize tracking error and control effort, using the bio-inspired Multi-Objective Particle Swarm Optimization (MOPSO) algorithm. Subsequently, to contrast the quality of the Pareto front obtained, the evolutionary Non-dominated Genetic Algorithm II (NSGA-II) was implemented, generating an expanded set of 26 non-dominated solutions. The comparative results showed that NSGA-II presents a better distribution of solutions and greater diversity of the Pareto front, while MOPSO offers faster convergence. Together, both approaches allow for more efficient tuning of the GPI controller, improving trajectory tracking by 23 % compared to traditional methods and suggesting the applicability of hybrid optimization strategies in the automation of agricultural robots.
Background With the increasing demand for logistics automation, the application of AGVs (Automated Guided Vehicles) in the industrial field become more widespread. Methods This paper presents a heavy-duty AGV with self-load/unload capability for 23 t sheet-metal stacks for BIW (body-in-white) press shops, elaborating its structural design, simulation, manufacturing and testing. Results The AGV enables autonomous movement and precise positioning of payload transfer without human intervention, which can enhance logistics efficiency, operational convenience and production safety. Finite element analysis was utilized to optimize the AGV structural strength and motion stability, while experiments validated the system feasibility and reliability. The developed AGV distinguishes itself through its high payload automation, robust structural design, and cost-effective precision positioning. Structural integrity under 3° shop-floor gradient and eccentric loading was verified by FEA. Buckling analysis of the upright under the eccentric-loaded condition returned a first-mode load multiplier of 92.8, far exceeding that of the ISO 3691-4:2020. Maximum stress and deflection under emergency braking case remain below 168 MPa and 1 mm allowable, respectively. Experimental results demonstrate that the AGV operates efficiently even under complex working conditions—docking trials achieved a mean error of 3.19 mm (specification ±5 mm). Conclusion the AGV satisfies all the static and dynamic strength requirements without any mass increase. This reduces labor costs and operational risks substantially and provides a practical solution for the automated upgrading of intelligent logistics equipment.
This study presents a legged mobile robot based on a regular polyhedral structure that achieves a point-symmetric body configuration. The proposed framework enables dynamic allocation of manipulator roles based on a single reference vector (e.g., gravity), allowing consistent functionality across different body orientations without posture-specific control. The approach was validated through simulations on three polyhedral structures (octahedral, dodecahedral, and icosahedral) and hardware experiments using an octahedral robot. Experimental results demonstrate that the robot can perform grasping and crouching behaviors in all six stable orientations, while locomotion is achieved in two orientations under the current actuator constraints, with an average locomotion speed of 0.092m/s across 10 independent trials. In addition, stable posture recovery was confirmed across 10 independent fall-recovery trials with an average recovery time of 2.26s. The results further showed that the same posture representation and role assignment algorithm could be consistently applied across different polyhedral geometries without modification of the control structure. These findings indicate that point-symmetric design provides a viable framework for achieving orientation-independent operation while maintaining a minimal and generalized control structure.
Post-stroke patients frequently adopt compensatory trunk movements to accomplish tasks. This can lead to the development of abnormal movement patterns, which can hinder functional recovery. To address this issue, this study proposes a Human-Centric assist-as-needed strategy based on a upper limb rehabilitation robot. By constructing a closed-loop human-machine system, the exoskeleton assistance force is dynamically adjusted to inhibit compensatory trunk movements in hemiplegic patients while ensuring task completion. It employs Long Short-Term Memory (LSTM) networks to predict trunk compensation angles (R 2 =0.99) and combines Convolutional Neural Networks with Bidirectional LSTM (CNN-BiLSTM) to calculate compensation torque (R 2 =0.97), thereby achieving personalized and active torque assistance. Eleven patients with upper limb hemiplegia participated in the clinical trial. The results indicated that trunk compensation during training was significantly improved in all patients. Human-centric assist-as-needed strategy significantly reduces the compensatory angles of hemiplegic patients in the sagittal plane (76.79%), coronal plane (75.82%), and horizontal plane (87.47%) (p < 0.01), outperforming both the no intervention and task-centric strategies, with a 100% task completion rate during patient training. Compared to task-centric strategy, human-centric assist-as-needed strategy enhances patient intention consistency (IC) by 31.91% and increases average torque (AT) output by 48.02%. This study represents the first application of deep learning in the suppression of active trunk compensation for upper limb exoskeletons, providing an innovative approach to stroke rehabilitation.
The design of a compact Gough-Stewart platform for physical hippotherapy is presented using pneumatic muscles and a central pre-stressed spring to achieve bidirectional motion. The use of the Gough-Stewart platform is advantageous as it drastically increases the workspace accessible by a therapist. The system is designed to reproduce walk, trot, and gallop trajectories derived from motion capture data. The model-based geometry optimization reduces actuator forces under workspace and collision constraints. The control of the pneumatic muscles combines inverse-dynamics and PID loops with a neural network feedforward. The latter compensates for the nonlinear spring characteristics. Experiments show that the system can generate the desired motion profiles, is robustness regarding variable load, and thus suitable for therapeutic use.
Surface electromyography (sEMG) is a noninvasive method for monitoring muscle activity, essential in rehabilitation, prosthetics, sports, and medicine. However, current sEMG systems struggle with signal noise, unclear muscle activation detection, and limited adaptability. This study proposes a method combining signal processing and machine learning to enhance muscle activation detection. Experiments on EMG data from muscle contractions ( n = 140 per test, with 28 participants, five biceps curl repetitions, five wrist curl repetitions each) demonstrated 92.23% overall accuracy (3.83% false positives, 3.94% false negatives) for biceps tests and 91.11% overall accuracy (4.36% false positives, 4.53% false negatives) for wrist tests, and an average recall of 96% significantly outperforming traditional methods. This approach highlights potential applications in real-world biomedical settings from rehabilitation and prosthetics to sports science.
This paper proposes an approach method that enables an autonomous underwater vehicle (AUV) to accurately approach an unmanned aerial vehicle (UAV)-deployed station in the final stage of a mission. AUVs are increasingly used as fully autonomous underwater survey platforms, yet their operation is still constrained by the need for human or vessel support during deployment and recovery. To remove this constraint, the UAV deploys an underwater station that provides a positioning reference and acts as a recovery system. Using onboard acoustic positioning and communication devices, the AUV and the station mutually transmit signals, allowing the AUV to estimate its state relative to the station and guide its approach. Sea experiments evaluated the resulting approach accuracy relative to the station, showing that the AUV can converge to the station vicinity with the precision required for subsequent recovery. This study validates only the pre-recovery approach phase to the recovery station. Full docking and recovery are beyond the scope of this article. The proposed method provides fundamental technology for a reliable close-range approach, which is a key prerequisite for fully autonomous recovery without human or vessel assistance.
The exponential growth of global data demands unprecedented reliability and efficiency of data center operations. Traditional manual inspection methods remain inefficient and error-prone. This article presents an autonomous mobile robot enhanced by large language models and hierarchical three-dimensional scene graphs for intelligent operation and maintenance. The core innovation lies in enabling semantic-aware navigation, allowing the robot to interpret high-level instructions like “inspect servers with the red alert light on in cold aisle B” by leveraging the three-dimensional scene graph for spatial-semantic reasoning and the large language model for task parsing. Deployed in multiple ultra-large data centers, the system has improved inspection efficiency by over 50%, autonomously patrolled over 10,000 km, and identified more than 1200 equipment anomalies. This work demonstrates a significant step towards fully autonomous, intelligent infrastructure management.
This paper presents a study on pedestrian detection and estimation from UAS imagery using recent YOLO-based object detection models. The objective is to evaluate model performance for identifying humans from aerial perspectives and to develop a customized detector suited for UAS applications. The study demonstrates the potential of combining modern artificial intelligence models with UAS-mounted vision systems for applications such as crowd monitoring, autonomous surveillance, and search-and-rescue operations. Experimental results demonstrated that the model achieved consistent detection accuracy up to 40m altitude, achieving near-perfect pedestrian identification with minimal false positives. The framework demonstrated its robustness for real-time deployment in aerial surveillance, search and rescue operations, and crowd monitoring scenarios.
With the advancement of intelligent manufacturing technology, there is an urgent demand for robots to replace humans in high-precision, heavy-duty, and other transportation tasks. In contrast to large-scale robotic arm manipulation, this work focuses on achieving suction-based grasping and transportation using an unmanned ground vehicle equipped with a small robotic arm. Compared with large manipulators, small robotic arms require higher precision and have lower fault tolerance. To improve exploration efficiency and sample utilization, this work proposes an enhanced multi-agent deep reinforcement learning method called curiosity-driven Multi-Agent Deep Deterministic Policy Gradient (CD-MADDPG). This method integrates a curiosity-driven prioritized experience replay mechanism into the multi-agent deep deterministic policy gradient framework, where the prediction residual of a forward dynamics model is used to quantify the novelty of samples, guiding the agent toward underexplored states. In addition, a decoupling strategy is adopted, where each joint of a single robotic arm is treated as an independent agent, thereby transforming the high-dimensional action space into a low-dimensional one, complemented by a designed global reward mechanism. Experimental results demonstrate that CD-MADDPG achieves a 19% improvement in success rate and a 44% improvement in distance accuracy compared to the non-decoupled single-agent counterpart, effectively accomplishing the suction-based grasping and transportation task.
Non-rigidly coupled LiDAR-camera systems are increasingly adopted for 3D reconstruction and autonomous robots. However, point cloud colorization remains challenging for such systems due to the absence of fixed extrinsic parameters between sensors. This paper proposes a novel colorization framework for self-built non-rigid systems, which decomposes the colorization task into localized 3D-2D projective transformation estimation, mitigating the lack of fixed extrinsics. Each image then colors local 3D scene within its view frustum. When merging these scenes, overlapping regions yield multiple coloring candidates for individual 3D points. To resolve this, we introduce a coloring reliability assessment method that selects optimal coloring sources per point by evaluating projection geometry and reprojection errors. We conduct comprehensive experiments on self-collected outdoor real-world datasets, reporting the quantitative accuracy of 3D-2D transformation estimation and conducting qualitative analysis of colorized point clouds; we also perform ablation studies on the key module and cross-system validation on rigidly coupled systems, all of which collectively demonstrate the effectiveness and extensibility of the proposed method.
Understanding the complex biomechanics of the ankle is essential for advancing rehabilitation protocols and improving the design of assistive devices. This work proposes a standardised and reproducible protocol for the three-dimensional assessment of ankle kinematics using a marker-based gold-standard motion capture system. Ankle movements of 34 healthy adult women and men were investigated using a VICON motion capture system equipped with 12 infrared cameras. A total of 32 reflective markers were attached to the lower legs, feet and shoes of each participant to capture joint motion and foot–shoe interactions. Participants performed controlled plantarflexion–dorsiflexion, inversion–eversion and abduction–adduction movements, which were analysed using combined local coordinate frames to resolve ankle motion in all three anatomical planes. The ankle joint exhibited the greatest mobility in the frontal plane, with a maximum pitch angle of 106.5; dorsiflexion and plantarflexion were dominated by pitch motion, with an average range of 59.1° in pitch; inversion and eversion showed a more evenly distributed motion pattern, with average angular displacements of 37.4° in roll; abduction and adduction were characterised by yaw motion, with 28.3°. Results demonstrated consistent intra-subject repeatability across trials, with noticeable inter-subject variability, confirming the effectiveness of the proposed protocol in capturing natural variations in human motion. A secondary analysis revealed relative displacement between the foot and shoe, with an average slip of approximately 1 mm and peak values exceeding 10 mm in extreme cases, highlighting the importance of footwear-foot coupling in kinematic studies. The proposed methodology provides a robust foundation for the quantitative characterisation of ankle mobility, enabling reproducibility across laboratories and supporting future developments in rehabilitation robotics, ergonomic footwear design and motion analysis research.
Recent research in unmanned system autonomy has focused on multi-robot systems, such as vehicles or unmanned aerial vehicles (UAVs), performing missions autonomously without mutual collisions in dynamic environments. In multi-agent operational settings, conventional centralized path planning methods face limitations in system scalability, as computational complexity increases sharply with the number of agents. Therefore, this research proposes a collision-aware adaptive horizon model predictive control (MPC) algorithm based on distributed model predictive control, which is advantageous for scalability. For computational efficiency, the proposed algorithm dynamically adjusts the length of the prediction horizon based on whether a collision is predicted on the planned path, and integrates a control barrier function (CBF) as a constraint to ensure safety even when the prediction horizon is shortened. The entire optimization problem is formulated as a computationally efficient quadratic programming; however, the linearized constraints used in this formulation can lead to deadlock. To address this issue, this work applies a detour strategy to increase the success rate of path planning. The performance of the proposed algorithm was validated in a three-dimensional simulation of a path-crossing scenario with a multi-agent system of UAVs. Specifically, an ablation study analyzing the computational efficiency of the adaptive horizon and the safety enhancement from the CBF demonstrated that the proposed method enables agents to reach their target destinations efficiently and safely, reducing the average computation time by 35% and increasing the mission success rate by 13%.
To improve the walking stability of a hexapod robot, a foot-end trajectory planning method using quintic polynomial interpolation through waypoints is presented. The main objective is to achieve smooth joint motion at waypoint transitions and to reduce vertical body oscillation during walking. First, an integrated model of the hexapod robot and a three-joint leg is established, and the kinematic relationship between joint angles and foot-end position is derived. Then, waypoint, stride, and continuity constraints are imposed on joint position, velocity, and acceleration to construct a general trajectory planning function based on quintic polynomial interpolation. Finally, an Adams–Simulink cosimulation model is developed, and the proposed method is verified under a triangular gait. The results indicate that the planned trajectory remains smooth when passing through waypoints, with no discontinuities in joint angular velocity or angular acceleration at transition points. Moreover, the robot achieves stable triangular-gait locomotion, while the vertical oscillation of the body is maintained within 1 mm.
Safe and efficient navigation for multi-robot systems in cluttered, dynamic environments remains challenging, primarily due to uncertainties from internal disturbances and dynamic external conditions. While existing methods enable real-time navigation in static or sparse settings, they often fail to respond effectively in environments dense with both static and moving obstacles. To address this limitation, this article proposes a novel safety probability field-based extended state model predictive control (SPF-EMPC) planner. The framework first introduces a safety probability field to model dynamic obstacles and integrates it with an unconstrained optimization approach for online generation of collision-free trajectories. Subsequently, an extended state model predictive controller ensures accurate trajectory tracking by explicitly accounting for robot model constraints and state perturbations, thereby guaranteeing practical feasibility. Both simulations and physical experiments demonstrate that the proposed method reliably prevents inter-robot and robot–obstacle collisions, even under significant motion and control uncertainties.
