Reconfigurable robots that alternate between compact rolling-related configurations and legged locomotion require feedback that is both observable and timely enough to supervise morphology changes. This study characterizes a spherical hexapod with 30 ZX20D serial-bus servos and one S80 center-of-mass (COM) shift actuator using the submitted 12.00 s synchronized record spanning HOME, FOLD, three COM-shift states, UNFOLD, and alternating tripod gait. Recalculation from the supplied 30-servo command/feedback archive yields statewise aggregate tracking RMSE values of 1.12–2.29 deg. The largest instantaneous errors are 33.51 deg at ROLL3-UNFOLD and 32.25 deg at a GAITB reversal. Under the timing assumptions, a modeled 30-servo scan requires 26.70 ms and corresponds to a nominal upper scan rate of 37.45 Hz. COM offsets of 72, 66, and 60 mm correspond to dominant-axis attitude peaks of 24.25, 21.55, and 20.02 deg. The uncalibrated contact/load-proxy channels separate the commanded stance and swing groups by a 53.0:1 mean ratio in this record. The resulting evidence is intended to define measurable prerequisites for prospective closed-loop supervision.
Autonomous driving (AD) has great potential for enhancing traffic safety and transportation efficiency. However, conventional AD methods often lack sufficient generalization and reasoning capabilities in dynamic and complex scenarios. Although large language models (LLMs) can alleviate this limitation, their inherent hallucination problem may introduce safety risks into decision making. To address these challenges, this paper proposes FEDrive-LLM, a feedback-enhanced closed-loop autonomous driving framework. This framework integrates the core modules of environment interaction, memory retrieval, reasoning decision, online safety verification, and reflection feedback. Furthermore, the safety verification module in FEDrive-LLM conducts online collision risk assessment for driving decisions, effectively mitigating the safety risks caused by model hallucinations. Meanwhile, the reflection feedback module reviews potentially conservative decisions and generates concise suggestions for subsequent steps, helping the system achieve a better balance between safety and driving efficiency. Simulation results show that FEDrive-LLM significantly outperforms existing state-of-the-art methods across multiple metrics, including success rate, success steps, and average speed. These results demonstrate its robustness and effectiveness in complex traffic environments.
The deployment of Unmanned Ground Vehicles in unstructured environments is currently polarized between lightweight scouting platforms and heavy industrial machinery, leaving a significant void in the functional capabilities of mid-sized robotic systems. This systematic review investigates the architectural constraints of the mid-size category, strictly defined in this work by a total Gross Vehicle Weight (GVW) ranging from 100 to 1000 kg through a Comparative Boundary Analysis of 24 empirical studies. By stratifying the literature into lightweight, mid-size, and heavy-duty clusters, the analysis identifies a Scalability Trap where kinematic principles effective at small scales become energetically prohibitive when upscaled. Specifically, the review demonstrates that the resistive turning moment inherent to skid-steering architectures scales linearly with mass, rendering differential drive mechanisms increasingly inefficient and prone to immobilization as payload scales in highly deformable soils. Furthermore, the synthesis of suspension dynamics reveals that rigid chassis designs in this intermediate class struggle to manage dynamic vertical forces, leading to high-frequency sensor noise and intermittent traction loss that pneumatic tyre compliance cannot mitigate. Algorithmically, the review establishes that reactive kinematic controllers frequently diverge under the inertial non-linearities of this mass regime during high-speed or transient maneuvers, necessitating predictive dynamic frameworks capable of explicitly managing actuator saturation and stochastic wheel slip. Despite these architectural and algorithmic limitations, empirical evidence confirms the agronomic superiority of the mid-size class, which delivers measurable crop yield improvements relative to conventional tractors by preserving soil porosity through reduced ground pressure. Consequently, this study concludes that the future of field robotics depends on shifting from rigid, skid-steered prototypes to morphologically adaptive systems utilizing 4-Wheel Steering, passive articulation, and predictive traction control to unlock the operational potential of the mid-size sector.
The human body has been extensively modeled for motion analysis and state estimation. In nursing care and workplaces, where large-scale measurement systems are impractical, a simpler system to measure the human body is required. Therefore, owing to their simplicity, two-dimensional human-link models in the sagittal plane are commonly used. If all link lengths are fixed, the model fails to compute certain postures owing to discrepancies with the actual human body. Hence, models have been proposed in which only the thigh or upper arm link length is variable, using changes in the apparent link length with human posture. The model can absorb errors by introducing variability in link lengths. The variable-length upper-arm model, which considers shoulder degrees of freedom, introduces relatively small errors but exhibits significant inaccuracies depending on posture. In this study, we analyzed human standing motion to propose a variable-length upper-body link model that maintains simplicity while reducing joint position errors across different postures. Because the upper body can flex, its apparent link length changes considerably, which helps reduce joint position errors. To validate the proposed model, we targeted users of standing support robots and compared the joint position errors obtained from both the proposed and existing models against actual positions. The standing up movements of eight young healthy individuals were measured, and the errors were compared using VisionPose® s the ground truth. The results of the t-test indicate that the model estimates joint positions with greater accuracy while being as simple as the existing models.
This paper presents a study on Human-Robot Teaming (HRT), focusing on state-of-the-art methodologies and real-world implementation of a collaborative system for industrial assembly tasks. The study explores the integration of real-time human action recognition using deep learning, specifically LSTM (Long Short-Term Memory) networks fed by upper-body skeletal data (limbs and torso) extracted via MediaPipe, in combination with object detection using the YOLO (You Only Look Once) method. The system allows a UR5e collaborative robot to dynamically adjust its assistance in response to the operator’s actions, specifically anticipating the tool needed based on the context rather than a fixed schedule. Unlike traditional automation, which relies on rigid, pre-programmed sequences, the proposed approach is sequence-independent. This flexibility allows the robot to support variable assembly workflows, with the potential to mitigate operator cognitive workload. The system enables the robot to autonomously select and offer the required tool to the operator at the right moment, further enhancing collaboration.
