Efficient waste detection is crucial for environmental sustainability, yet existing models struggle with finegrained objects in complex backgrounds, such as those in the TACO dataset. This paper proposes an attention-centric approach using YOLOv12n to balance detection accuracy and real-time performance. Experimental results on the TACO dataset demonstrate that the proposed YOLOv12n achieves a mean Average Precision (mAP50 ) of 0.376 with an end-to-end inference speed of 94.33 FPS on an NVIDIA RTX 5060 GPU. Ablation studies reveal that removing the Area-Attention ( A2 ) module leads to a significant performance drop, with mAP50 plummeting from 0.376 to 0.148. Furthermore, compared to the YOLOv8n model with a plug-in CBAM module (46.18 FPS, 0.310 mAP), the native attention-centric architecture of YOLOv12n provides a significant increase in inference speed and superior feature localization. This research confirms that a native attention-based design is more effective for realtime fine-grained waste detection than traditional modular additions.

This paper reviews the latest research progress and key technologies in accelerating edge computing with optical neural networks, set against the backdrop of autonomous driving. It first outlines the context and research significance of autonomous driving, highlighting the technical limitations faced by traditional electronic computing. Subsequently, it provides a detailed analysis of the latest research achievements in optical neural networks and accelerated computing both domestically and internationally. This includes innovative applications of optical flow estimation methods (FocusFlow, CSFlow), introductions to methods for edge computing and autonomous driving collaboration (OVEAP, DVEAP), and innovative approaches to underlying computations (optical convolution, Monet channels). Through comparative analysis, it reveals current re-search hotspots and trends, analyzes existing short-comings and potential solutions, and explores possible future development directions. This paper aims to systematically explore how optical computing technology can overcome the limitations of traditional electronic computing to build a new generation of high-performance, low-power intelligent computing systems.

This paper explores the battery energy storage systems (BESS) usage to achieve substantial voltage stability in residential micro grids with high photovoltaic (PV) penetration. It is motivated by the growing need to use inverter-based resources, which, in contrast to synchronous generators, offers less inertia and reactive support, which results in such issues as voltage sag, swells, and poor regulation during disturbances. A dynamic model of a residential micro grid comprising of five nodes was modeled in equation-based MATLAB, expressed as a differential-algebraic system to be solved by means of the Newton-Raphson iteration and backward integration with the Euler method. Load steps, PV generation ramps, short-circuit faults and grid islanding were all considered disturbance scenarios. Three types of inverter control schemes, namely grid-following PQ mode, droop-based support, and emulation of a virtual synchronous generator (VSG) were tested. The simulated cases with and without BESS were done in several replicates to provide statistical reliability. The outputs of primary data were bus voltages, active and reactive flows of power, and state-of-charge (SoC). Measures of stability were established as the index of voltage deviation, the percentage of time below the +-5% statutory band, minimum bus voltage and a Jacobian singular value proxy of margin evaluation. The findings also indicate that consistent implementation of BESS decreases the variability of the peak voltage and decreases the range of variability of all disturbances. As an example, BESS minimized mean deviation by almost 40 per cent under PV ramp conditions and by almost 50 per cent during load step events. Control grid strategies (droop, VSG) were better at giving transient stability than PQonly operation, VSG being smoother to recover than droop (because of virtual inertia). Sensitivity tests indicated that BESS and state-of-charge flexibility have a significant impact on performance, and the reduction in returns with increasing BESS capacity is steep: beyond approximately 30 per cent of peak load rating. These results reaffirm that BESS on residential scale, even small-scale, can significantly enhance stability of voltage, which is of benefit to engineering design and policy projects aimed at resilient PV-heavy microgrids. The article offers a clear and reproducible simulation model and indicates further developments with hardware-in-theloop testing, unbalanced three-phase modeling, and technoeconomic optimization.

In the past 10 years, the evolution path of medical imaging AI in terms of model architecture is very clear: it is evolving from CNNs for end-to-end feature learning, to ResNets for alleviating deep degradation issues, to EfficientNet balancing accuracy and efficiency with compound scaling, to Transformers for global modeling with self-attention. It is systematic and clear, but it is also important to map it systematically, which will help to be clearer about technology selection and further optimize clinical translation. This paper focuses on four major models/captures CNN, ResNet, EfficientNet, and Transformer, summarizes their basic designs and innovation points.

Concern around global traffic accidents have increased due to the large toll they take on people and the economy. To investigate factors affecting the result of vehicle accidents in the UK, this study attempts to predict the severity of accidents using a Random Forest model. It investigates difference feature sets including time, context and interaction. Following data preprocessing and class balancing using the Synthetic Minority Oversampling Technique (SMOTE), five feature sets (baseline, temporal, context, interaction and extended) were trained and assessed over 10 iterations. The results show that overall accuracy remains around 0.78 and AUC around 0.60 across all models, indicating stable performance. However, the inclusion of contextual and interaction features slightly improves recall for fatal and serious crashes and reduces the dominance of a few strong predictors, leading to a more balanced feature contribution. These results suggest that richer feature design enhances Random Forest’s ability to capture complex crash patterns beyond the majority of slight accidents.

Against the backdrop of intensifying global climate change and the deepening advancement of carbon neutrality strategies, carbon capture, utilization, and storage (CCUS) technology has emerged as a critical deep decarbonization approach, increasingly capturing the attention of the international community. This paper systematically reviews the current development status of the entire CCUS technology chain, covering the classification and maturity levels of technologies across capture, transportation, utilization, and storage segments. Drawing on international exemplary projects and China’s development practices, this analysis examines the multifaceted challenges hindering the large-scale application of CCUS, including economic costs, environmental risks, and policy mechanisms. Although the CCUS technology pathway is gradually becoming clearer and its application scale is expanding, core issues such as high energy consumption and costs, insufficient system integration, and an incomplete institutional framework still need to be overcome. Finally, this paper proposes a technology roadmap centered on reducing energy consumption and improving efficiency, along with policy recommendations involving carbon market mechanisms and cross-regional cooperation. These suggestions provide a theoretical reference and decisionmaking basis for advancing CCUS commercialization and supporting carbon neutrality goals.

The ocean serves as a vast and largely underexplored frontier, playing a critical role in global resource provision and climate regulation. In recent years, as understanding of its ecological and economic significance has deepened, underwater robots, particularly Autonomous Underwater Vehicles (AUVs), have emerged as indispensable tools for deep-sea exploration and development. Their unique capability to operate untethered from surface vessels offers distinct advantages over Remotely Operated Vehicles (ROVs), including a wider operational range, superior depth tolerance, enhanced maneuverability, and lower maintenance demands. With the growing strategic importance of the oceans in the 21st century, AUVs have proven vital in diverse fields such as resource assessment, scientific research, and military operations. This review aims to synthesize the current state of AUV control technology, identify pivotal research challenges, and outline emerging future trends. It examines a spectrum of control methodologies, ranging from conventional approaches to cutting-edge intelligent algorithms, while focusing on the unique constraints and testing challenges posed by the complex and unpredictable underwater environment.

The rapid development of renewable energy and batteries has increased people’s demand for lithium-ion batteries. Thermal management plays a very important role in the manufacturing of lithium-ion batteries, as overheating is an inevitable problem when batteries are used for a long time. Moreover, thermal. However, during the charging process, the battery will generate a large amount of heat. This excessive heat will cause the battery’s capacity to decrease and also pose a risk of heat runaway. It also plays a crucial role in ensuring the safety, efficiency and lifespan of the battery. This article, however, reviews the latest advancements in thermal conductive materials applied in battery thermal management systems. Using the CAS method, this study systematically analyzed the advantages and disadvantages of carbon-based and metalbased related materials, and particularly emphasized their thermal conductivity and stability. Furthermore, this article also discusses the technologies that enhance conductivity while also reducing costs, as well as the innovations related to this technology, and the future development trends of composite materials. By comparing different materials, this study ultimately concluded that although there is no single material type that can completely solve the problem of thermal management, the use of hybrid materials and new materials still holds potential for the sustainable development of lithium-ion technology.

The precise operation of minimally invasive surgical (MIS) robots relies on multi-dimensional perception of the surgical scene, and the fusion of force feedback (tactile perception) and visual perception is the core technology to break the bottleneck of the “perception-decision-control” loop. This paper systematically reviews the development of force-visual fusion technology for MIS robots, analyzing from sensor layer innovation, and algorithm layer breakthroughs, to system layer integration. It examines technical adaptability and performance bottlenecks by combining typical clinical scenarios such as tumor resection, vascular suturing and nerve dissection. Finally, addressing key issues including poor generalization in in-vivo environments, difficulties in multi-modal spatiotemporal synchronization, and lack of clinical translation standards, the paper proposes future development paths based on “surgical metaverse,” “hybrid dataset training,” and tele-surgery specifications, providing references for technology R&D and clinical translation in the interdisciplinary field of medicine and engineering.

Due to the rapid development of artificial intelligence, prompt engineering technology has become a key technology for the development of Large Language Models in the field of generation tasks. Although the existing reviews have listed various technologies in detail, they still lack an analysis of the inherent logic and evolution laws among the technologies. Therefore, this article classifies the core issues that the technology solves, namely the functional goals, establishes the connection between the technology and the core issues, and presents the development path of the technology. First, let’s introduce the basic prompt class to achieve something from nothing and solve the problem of task understanding. Then elaborate on how structured reasoning classes deal with logical reasoning problems. In this category, it is further divided into two subcategories: linear reasoning and nonlinear reasoning. Then, it discusses how the hallucination-correcting category overcomes the problems of knowledge deficiency and hallucinations. Finally, it explores how the automation optimization category can focus on the efficiency of manual design and promote the evolution of technology towards scale and industrialization. In the future, engineering technology will move towards a more intelligent direction, achieving a paradigm shift in human-computer interaction.