Frontiers in Robotics and AI最新文献

Robot applications encompass a multitude of edge computing tasks, such as image processing, health monitoring, path planning, and infotainment. However, task scheduling within such environments remains a significant challenge due to the inherent limitations of edge computing resources and the dynamically fluctuating nature of workloads. EdgeCloudSim, a widely used simulation platform for edge computing, supports a conventional control strategy-Least-Loaded First-Fit Decreasing (LLFFD)-that is favored for its simplicity and speed, especially in scenarios with relatively small-scale and stable workloads. However, as the number of tasks grows and task-VM matching becomes more complex, traditional heuristics struggle to optimize resource utilization and energy consumption effectively. To address this, we propose a hybrid scheduling approach-FFDDE-that integrates the FFD heuristic with the Differential Evolution (DE) algorithm for optimized task-to-VM mapping in edge environments. Using the EdgeCloudSim simulation framework, we evaluate both strategies under diverse workload conditions, comparing their performance in terms of energy consumption and task completion time. Experimental results demonstrate that, compared with the traditional LLFFD method and the classic heuristic algorithm-GA, the hybrid DE-based strategy achieves significantly improved energy efficiency through better task consolidation. This study highlights the potential of combining fast heuristic methods with evolutionary optimization to achieve more sustainable task scheduling in edge computing scenarios.

Real-time estimation of human action progress is critical for seamless human-robot collaboration yet remains underexplored. With this paper we propose the first real-time application of Open-end Soft-DTW (OS-DTW_EU) and introduce OS-DTW_WP, a novel DTW variant that integrates a Windowed-Pearson distance to effectively capture local correlations. This method is embedded in our Proactive Assistance through action-Completion Estimation (PACE) framework, which leverages reinforcement learning to synchronize robotic assistance with human actions by estimating action completion percentages. Experiments on a chair assembly task demonstrate OS-DTW_WP's superiority in capturing local motion patterns and OS-DTW_EU's efficacy in tasks presenting consistent absolute positions. Moreover we validate the PACE framework through user studies involving 12 participants, showing significant improvements in interaction fluency, reduced waiting times, and positive user feedback compared to traditional methods.

Introduction: Robotics and wearable systems are increasingly being discussed as potential solutions to address the physical demands, skill shortages and safety risks faced by the construction industry. However, their successful implementation hinges not only on technical feasibility, but also on their alignment with real working conditions. This article examines how interactive robotic systems and exoskeletons are experienced by construction workers by integrating macro-level data from European and national surveys with micro-level insights from pilot studies.

Methods: Five large-scale European surveys were analysed and combined with data from four pilot studies involving 37 workers interacting with three robotic prototypes and one upper-body exoskeleton. Quantitative data included usability, workload, interaction principles and affinity for technology. Qualitative feedback was obtained through open-ended responses.

Results: A set of guidelines for a human-centred approach to inform policy were derived, offering practical guidance on designing and deploying interactive robotic systems that are functional, safe, acceptable and effective in changing work environments.

Discussion: The observed challenges highlight the gap between the early stages of system design and the realities of dynamic construction work, emphasising the need for a participatory, human-centred development approach. The findings suggest that a human-centred approach is essential for emerging technologies to be functional, safe, acceptable and effective in changing work environments.

Swarm perception enables a robot swarm to collectively sense and interpret the environment by integrating sensory inputs from individual robots. In this study, we explore its application to people re-identification, a critical task in multi-camera tracking scenarios. We propose a decentralized, feature-based perception method that allows robots to re-identify people across different viewpoints. Our approach combines detection, tracking, re-identification, and clustering algorithms, enhanced by a model trained to refine extracted features. Robots dynamically share and fuse data in a decentralized manner, ensuring that collected information remains up to date. Simulation results, measured by the cumulative matching characteristics (CMC) curve, mean average precision (mAP), and average cluster purity, show that decentralized communication significantly improves performance, enabling robots to outperform static cameras without communication and, in some cases, even centralized communication. Furthermore, the findings suggest a trade-off between the amount of data shared and the consistency of the Re-ID.

Imitation learning (IL) has fundamentally transformed the field of legged robot locomotion, removing the dependence on hand-engineered reward functions. Since 2019, this area of research has progressed rapidly, from simple motion-capture replication to the generation of sophisticated policies using diffusion models. This survey offers a comprehensive analysis of 35 pivotal research works, using a structured six-dimensional framework to investigate advancements using quadrupedal and humanoid platforms. The review also pinpoints significant challenges related to deployment and outlines new research directions. A key finding from the survey indicates that behavior cloning is utilized in almost half of the analyzed studies. Moreover, data generated through model-predictive control (MPC) now represents the most frequently used training data source for advanced imitation learning systems.

Introduction: In mining regions of Latin America, thousands of children and adolescents are deprived of formal education because of their participation in labor-intensive economic activities. This study addresses how educational robotics can serve as a strategy for both social inclusion and pedagogical intervention in communities with disrupted or nonexistent schooling.

Methods: A multi-site intervention was implemented, directly benefiting 2,500 out-of-school or at-risk youth and 250 teachers in rural mining regions. The initiative encompasses the design and construction of educational robots and learning materials by university engineering students. Activities were conducted via project-based learning sessions and teacher training workshops. A mixed-methods approach was employed, integrating surveys, interviews, and participant observation to assess the impact on motivation, re-engagement with schooling, and pedagogical practices.

Results: The findings indicated increased student engagement, enhanced collaborative learning, and a measurable rise in school re-enrollment within the participating communities. Educators reported enhanced confidence in utilizing technological tools and heightened motivation among students. The robots acted as mediating artifacts, facilitating dialogical, hands-on learning experiences and bridging gaps between formal education and local realities.

Discussion: The results underscore the potential of educational robotics to serve not just as a pedagogical instrument but also as a transformative vehicle for fostering inclusion, motivation, and equity in marginalized environments. The initiative also demonstrates the significance of university-community collaboration in addressing educational inequality through innovation. Challenges include maintaining long-term impact and scaling the model to other contexts with similar vulnerabilities.

Detecting surface discontinuities in welds is essential to ensure the structural integrity of welded elements. This study addresses the limitations of manual visual inspection in shielded metal arc welding by applying convolutional neural networks for automated discontinuities detection. A specific image dataset of discontinuities on Shielded Metal Arc Welding weld seams was developed through controlled experiments with various electrode types and welder experience levels, resulting in 3,000 images. The YOLOv7 architecture was trained and evaluated on this dataset, achieving a precision of 97% and mAP@0.5 of 94%. Results showed that increasing the dataset size and training periods significantly improved detection performance, with optimal accuracy observed around 250-300 epochs. The model demonstrated robustness to moderate variations in image aspect ratio and generalization capabilities to an external dataset. This paper presents an approach for detecting SMAW weld surface discontinuities, offering a reliable and efficient alternative to manual inspection and contributing to the advancement of intelligent welding quality control systems.

Rehabilitation robots are widely recognized as vital for restoring motor function in patients with lower-limb impairments. A Modified Fractional-Order Proportional-Integral-Derivative (MFOPID) controller is proposed to improve trajectory tracking of a 2-DoF Lower Limb Rehabilitation Exoskeleton Robot (LLRER). The classical FOPID is augmented with a modified control formulation by which steady-state error is reduced and the transient response is sharpened. Controller gains and fractional orders were tuned offline using a hybrid metaheuristic Improved Elk Herd Optimization hybridized with Grey Wolf and Multi-Verse Optimization algorithms (IElk-GM) so that exploration and exploitation are balanced. Superiority over the classical FOPID was demonstrated in simulations under linear and nonlinear trajectories, with disturbances and parametric uncertainty: 0% overshoot was achieved at both hip and knee joints; settling time was reduced from 6.998 s to 0.430 s (hip) and from 7.150 s to 0.829 s (knee); ITAE was reduced from 23.39 to 2.694 (hip) and from 16.95 to 3.522 (knee); and the hip steady-state error decreased from 0.018 Rad to 0.0015 Rad, while the knee steady-state error remained within 0.011 Rad. Control torques remained bounded under linear tracking (<345 N·m at the hip; <95 N·m at the knee) and under nonlinear cosine tracking (<350 N·m at the hip; <100 N·m at the knee). These results indicate that safer, smoother, and more effective robot-assisted rehabilitation can be supported by the proposed controller.