Communications engineering最新文献

Chemical exchange saturation transfer (CEST) is a promising magnetic resonance imaging (MRI) technique that provides molecular-level information in vivo. To obtain this unique contrast, repeated acquisition at multiple frequency offsets is needed, resulting a long scanning time. In this study, we propose a hybrid strategy at k-space and image domain to accelerate CEST MRI to facilitate its wider application. In k-space, we developed a complementary undersampling strategy which enforces adjacent frequency offsets by acquiring different subregions of k-space. Both Cartesian and spiral k-space trajectories were applied to validate its effectiveness. In the image domain, we developed a multi-offset transformer reconstruction network that uses complementary information from adjacent frequency offsets to improve reconstruction performance. Additionally, we introduced a data consistency layer to preserve undersampled k-space and a differentiable coil combination layer to leverage multi-coil information. The proposed method was evaluated on rodent brain and multi-coil human brain CEST images from both pre-clinical and clinical 3 T MRI scanners. Compared to fully-sampled images, our method outperforms a number of state-of-the-art CEST MRI reconstruction methods in both accuracy and image fidelity. CEST maps, including amide proton transfer (APT) and relayed nuclear Overhauser enhancement (rNOE), were calculated. The results also showed close agreement with fully-sampled ones.

A parabolic coast or wall concentrates incoming waves at its focal point, creating a high‑energy zone ideal for enhanced capture. Yet, how to efficiently harvest this concentrated energy remains unclear. Here we propose designs of single- and dual-chamber Oscillating Water Column (OWC) chambers for enhancing wave energy capture. A time‑domain higher‑order boundary element method, grounded in nonlinear potential flow theory, is coupled with a nonlinear pneumatic model-calibrated via geometric scaling, dual‑chamber coupling, and focused‑wave boundary tests-to simulate OWC performance. Under parabolic focusing, a bimodal resonance yields peak power absorption up to 17 times that of an isolated device, and a leeward perforation design boosts the single‑chamber capture ratio to 25 times baseline. A dual‑chamber configuration with an added semicircular chamber further elevates total absorbed energy and widens the effective bandwidth. This work provides practical design guidance for efficient wave-energy devices operating in focused-wave environments.

Distillation is the most energy-consuming unit operation of the chemical industry, however, its decarbonization strategy necessitates laborious manual process simulation, optimization and carbon emission accounting. Here we established a reasoning agent consisting of a large language model (LLM) and an extensive tool set to automate learning material collection, process simulation, optimization and carbon emission accounting of a representative methanol and ethanol distillation case study. Then the agent automatically constructed a heat pump-assisted distillation process to save energy. The impact of three energy supply scenarios on the carbon emissions of distillation, namely, coal, natural gas and renewables, was evaluated. Combining the heat pump-assisted process and renewables could substantially reduce the carbon emission by 98% compared with the coal-based traditional distillation process. This study explored using reasoning agents to automate carbon emission and decarbonization intervention quantification, and facilitated high-resolution carbon emission models of the industry.

Truck platooning promises to enhance the efficiency of logistics, but commercial operation is hampered by safety and economic concerns. Human-lead truck platooning can mitigate these challenges by leveraging a human driver's expertise. However, existing human-lead truck platooning is limited to longitudinal control and lacks the lane-changing capability, which restricts logistical efficiency. To address this, we build upon previous research to propose a human-lead truck platooning method with lane-changing capability. The platoon leader is controlled by a skilled human driver, who is responsible for leading the following automated trucks. The human-lead platoon is enabled to cruise, lane-change, and obstacle avoidance, leveraging the driver's expertise to mitigate safety risks in long-tail scenarios. Drivers of the following trucks are not needed, reducing labor costs. The proposed method has been implemented in commercial operations at the world's largest port, Shanghai Yangshan Port, achieving an annual transport volume of 200,000 Twenty-foot Equivalent Units. It highlights a route for large-scale truck platooning implementation, potentially reshaping freight-transport operations.

Large-volume 3D dense prediction is essential in industrial applications like energy exploration and medical image segmentation. However, existing deep learning models struggle to process full-size volumetric inputs at inference due to memory constraints and inefficient operator execution. Conventional solutions-such as tiling or compression-often introduce artifacts, compromise spatial consistency, or require retraining. Here we present a retraining-free inference optimization framework that enables accurate, efficient, whole-volume prediction without performance degradation. Our approach integrates operator spatial tiling, operator fusion, normalization statistic aggregation, and on-demand feature recomputation to reduce memory usage and accelerate runtime. Validated across multiple seismic exploration models, our framework supports full size inference on volumes exceeding 1024³ voxels. On FaultSeg3D, for instance, it completes inference on a 1024³ volume in 7.5 seconds using just 27.6 GB of memory-compared to conventional inference, which can handle only 448³ inputs under the same budget, marking a 13 × increase in volume size without loss in performance. Unlike traditional patch-wise inference, our method preserves global structural coherence, making it particularly suited for tasks inherently incompatible with chunked processing, such as implicit geological structure estimation. This work offers a generalizable, engineering-friendly solution for deploying 3D models at scale across industrial domains.

Swarm control of autonomous underwater vehicles (AUVs) has been recognized as the foundation for marine exploration. However, the implementation of this task faces two major constraints: excessive communication energy demands and limited environmental perception capabilities. This article proposes a digital twin (DT)-driven swarm control of AUVs solution to overcome these limitations. We first create the digital replicas for each AUV by integrating the dynamics and environmental data. With the collected states from AUVs, a parameter estimator is proposed to predict the flow field, while a swarm networking protocol is designed to reduce the energy consumption. After that, an integral reinforcement learning (IRL)-based swarm controller is proposed to drive the virtual and real AUVs. Based on the interaction information between DT models and AUVs, the virtual-real error optimization algorithm is implemented to minimize the matching errors. Finally, the effectiveness of our solution is verified by the experimental results. These results demonstrate that the DT-driven swarm control of AUVs can improve the underwater situation awareness and prediction accuracy while reducing the communication energy consumption.

Using human responses to optimize and thus personalize assistance enhances exoskeleton performance during locomotion. Current approaches lack efficiency, comfort, rapid deployability, and computation and actuation simplicity. Here we present a method that optimizes assistance within 2 min, 16 times faster than the state-of-the-art, by effectively imitating human joint moment while ensuring stability. Optimization of a unilateral ankle exoskeleton with off-board actuation produced gentler assistance (78.2% torque) while reducing muscle activity by 36.8% and metabolic cost by 20.4% than no assistance, comparable to state-of-the-art. The method was easily and effectively deployed across new gait conditions, to bilateral devices, to knee joints and also outdoors. It largely avoided the problems of existing methods with instantaneously measurable feedback, a non-aggressive tuning process, a reasonable tuning direction, and a non-parametric assistance formulation. By significantly reducing pre-research, operational, user physiological and psychological costs, this method largely elevates the accessibility level of effective, personalized and continuously tuned exoskeletons in everyday scenarios.

Advances in adaptive optics optical coherence tomography (AOOCT) have facilitated the three-dimensional assessment of structural and functional properties of individual retinal cells in the living human eye. However, even with diffraction-limited AOOCT systems, some cells in the living human retina can be difficult to resolve, especially when using near-infrared wavelengths of light (~1000 nm). We demonstrate that modifying the traditional AOOCT instrument design to enable annular illumination and sub-Airy disk detection results in improved imaging resolution beyond fundamental limits imposed by diffraction. We successfully applied this approach to in vivo human retinal imaging, achieving on average 36% improvement in lateral resolution beyond conventional imaging conditions, enabling improved visualization of the foveal cone and rod photoreceptor mosaics using AOOCT. These results demonstrate an effective strategy for improving lateral resolution in point-scanning AOOCT in a manner that is compatible with new and existing instruments.

Underwater object detection plays a crucial role in applications such as marine ecological monitoring and underwater rescue operations. However, challenges such as limited underwater data availability and low scene diversity hinder detection accuracy. In this paper, we propose the Underwater Layout-Guided Diffusion Framework (ULGF), a diffusion model-based framework designed to augment underwater detection datasets. Unlike conventional methods that generate underwater images by integrating in-air information, ULGF operates exclusively on a small set of underwater images and their corresponding labels, requiring no external data. We have publicly released the ULGF source code and the generated dataset for further research. Our approach enables the generation of high-fidelity, diverse, and theoretically infinite underwater images, substantially enhancing object detection performance in real-world underwater scenarios. Furthermore, we evaluate the quality of the generated underwater images, demonstrating that ULGF produces images with a smaller domain gap.

Traditional plastic and glass culture lacks physiological relevance, undermining predictive power in drug discovery. Organoids and organs-on-chip improve biomimicry but do not scale to high-throughput screening (HTS). Even simple hydrogel coatings in HTS plates suffer from curved menisci that disrupt seeding and imaging. We present HYDRA (HYDrogels by Robotic liquid-handling Automation), an automated method to fabricate thin, planar hydrogel films directly in standard plates. Liquid-handlers dispense sub-contact volumes without wall wetting; immediate re-aspiration pins the contact line, leaving a uniform layer with controlled stiffness and thickness. Using fish gelatin hydrogel, HYDRA produces meniscus-free coatings compatible with routine 96- and 384-well workflows and plate-scale quality control. HYDRA was validated through imaging-based dose-response assays with anticancer compounds, engineered epithelial monolayers, and long-term holographic and fluorescence microscopy. It preserved pharmacological sensitivity while supporting high-content imaging on soft, biomimetic substrates, offering a practical bridge between physiological relevance and HTS scalability for early in-vitro drug testing.