A scalable approach to optimize traffic signal control with federated reinforcement learning.

Intelligent Transportation has seen significant advancements with Deep Learning and the Internet of Things, making Traffic Signal Control (TSC) research crucial for reducing congestion, travel time, emissions, and energy consumption. Reinforcement Learning (RL) has emerged as the primary method for TSC, but centralized learning poses communication and computing challenges, while distributed learning struggles to adapt across intersections. This paper presents a novel approach using Federated Learning (FL)-based RL for TSC. FL integrates knowledge from local agents into a global model, overcoming intersection variations with a unified agent state structure. To endow the model with the capacity to globally represent the TSC task while preserving the distinctive feature information inherent to each intersection, a segment of the RL neural network is aggregated to the cloud, and the remaining layers undergo fine-tuning upon convergence of the model training process. Extensive experiments demonstrate reduced queuing and waiting times globally, and the successful scalability of the proposed model is validated on a real-world traffic network in Monaco, showing its potential for new intersections.

Biomolecular 1D Necklace-like Nanostructures Tailoring 2D Janus Interfaces for Controllable 3D Enteric Biomaterials.

Construction of well-ordered two-dimensional (2D) and three-dimensional (3D) assemblies using one-dimensional (1D) units is a hallmark of many biointerfaces such as skin. Mimicking the art of difunctional properties of biointerfaces, which skin exhibits as defense and shelter materials, has inspired the development of smart and responsive biomimetic interfaces. However, programming the long-range ordering of 1D base materials toward vigorous control over 2D and 3D hierarchical structures and material properties remains a daunting challenge. In this study, we put forward construction of 3D enteric biomaterials with a two-strata 2D Janus interface assembled from self-adaptation of 1D protein-polysaccharide nanostructures at an oil-water interface. The biomaterials feature a protein dermis accommodating oil droplets as a reservoir for bioactive compounds and a polysaccharide epidermis protecting them from gastric degradation. Furthermore, the epidermis can be fine-tuned with different thicknesses rendering enteric delivery of a bioactive cargo (coumarin-6) with controllable retention in the intestinal tract from 6 to 24 h. The results highlight a skin-inspired construction of enteric biomaterials by self-adaptation of 1D nanostructures at the oil-water interface toward 2D Janus biointerfaces and 3D microdevices, which can be tailored for intestinal treatments with intentional therapeutic efficacies.

Diffusion-mediated carving of interior topologies of all-natural protein nanoparticles to tailor sustained drug release for effective breast cancer therapy.

Proteins are promising base materials for developing drug carriers with efficient blood circulation due to low possibilities of clearance by macrophages. However, such natural biopolymers have highly sophisticated molecular structures, preventing them from being assembled into nano-platforms with manipulable payload release profiles. Here, we report the self-assembly of two natural proteins (milk casein and rice protein) into protein nanoparticles (NPs, ∼150 nm) with tailorable release profiles. Diffusion of plant-derived paclitaxel (PTX)-containing eugenol into the hydrophobic cores of the NPs and subsequent dialysis to remove eugenol from the cores lead to the carving of the NP interiors. With the increase in the mass ratios of casein and rice protein, this process generates all-natural NPs with PTX loaded in their full cavities, semi-full cavities, or solid cores. These NPs can be efficiently uptaken by breast cancer cells and could kill the cancer cells efficiently. PTX in these NPs demonstrates increasingly sustained in vivo release profiles from full cavities, semi-full cavities, to solid cores, gradually extending its pharmacokinetic profiles in blood plasma to favor drug accumulation in breast tumor models. Consequently, the NPs with solid cores completely inhibit tumor growth in vivo, more effectively than those with full and semi-full cavities. Our work opens up a new avenue to the use of diffusion-mediated nanoscale carving in producing biomaterials with controllable interior topologies relevant to drug release profiles.

Efficient Estimation of Permeate Flux of Asymmetric Ceramic Membranes for Vacuum Membrane Distillation.

Ceramic membranes have the advantages of high mechanical strength and thermal stability and are promising candidates for membrane distillation. Ceramic membranes are generally designed to have a multilayer structure with different pore sizes to create a high liquid entry pressure and obtain a high permeability. However, these structural characteristics pose significant difficulties in predicting permeate flux in a ceramic membrane contactor for vacuum membrane distillation (VMD). Here, a modeling approach was developed to simulate the VMD process and verified by comparing the simulated results with the experimental data. Furthermore, correlations are proposed to simplify the calculations of permeate flux for VMD using asymmetric ceramic membranes by assuming those multilayers to be an effectively quasi-symmetric layer and by introducing a correction factor. The simulation results indicated that this simplified correlation was effective and enabled a quick estimation of the effect of membrane parameters on permeate flux.

Enhanced Performance of Fly Ash-Based Supports for Low-Cost Ceramic Membranes with the Addition of Bauxite.

Support is a necessary foundation for ceramic membranes to achieve high performance. Finding the optimum balance between high performance and low cost is still a significant challenge in the fabrication of ceramic supports. In this study, low-cost fly ash-based ceramic supports with enhanced performance were prepared by the addition of bauxite. The pore structure, mechanical strength, and shrinkage of fly ash/bauxite supports could be tuned by optimizing the bauxite content and sintering temperature. When the sintering temperature and bauxite content were controlled at 1300 °C and 40 wt%, respectively, the obtained membrane supports exhibited a high pure water permeance of approximately 5.36 m3·m-2·h-1·bar-1 and a high bending strength of approximately 69.6 MPa. At the same time, the optimized ceramic supports presented a typical mullite phase and excellent resistance to acid and alkali. This work provides a potential route for the preparation of ceramic membrane supports with characteristics of low cost and high performance.

Permeability and Stability of Hydrophobic Tubular Ceramic Membrane Contactor for CO2 Desorption from MEA Solution.

Ceramic membrane contactors hold great promise for CO2 desorption due to their high mass transfer area as well as the favorable characteristics of ceramic materials to resist harsh operating conditions. In this work, a hydrophobic tubular asymmetric alpha-alumina (α-Al2O3) membrane was prepared by grafting a hexadecyltrimethoxysilane ethanol solution. The hydrophobicity and permeability of the membrane were evaluated in terms of water contact angle and nitrogen (N2) flux. The hydrophobic membrane had a water contact angle of ~132° and N2 flux of 0.967 × 10-5 mol/(m2∙s∙Pa). CO2 desorption from the aqueous monoethanolamine (MEA) solution was conducted through the hydrophobic tubular ceramic membrane contactor. The effects of operating conditions, such as CO2 loading, liquid flow rate, liquid temperature and permeate side pressure, on CO2 desorption flux were investigated. Moreover, the stability of the membrane was evaluated after the immersion of the ceramic membrane in an MEA solution at 373 K for 30 days. It was found that the hydrophobic α-Al2O3 membrane had good stability for CO2 desorption from the MEA solution, resulting in a <10% reduction of N2 flux compared to the membrane without MEA immersion.

Ultrasound Assisted Synthesis of Size-Controlled Aqueous Colloids for the Fabrication of Nanoporous Zirconia Membrane.

Permeation and separation efficiency of ceramic membranes are strongly dependent on their nanoporous structures, especially on the pore size. In this work, ultrasound is employed to form the size-controlled ZrO2 nanoparticles, and a ceramic membrane is prepared with tunable pore size. Under the ultrasound treatment, H+ from water plays a key role in the synthesis process. The cavitation caused by ultrasound promotes the hydrolysis of the precursor in water, which produces a large number of H+. These H+ will react with precipitant added and generate cyclic tetrameric units. Excess H+ can peptize cyclic tetrameric units and form an electrical double layer, resulting in a stable sol. Unlike ultrasound treatment, precipitant will react directly with the precursor and generate precipitation if there is no ultrasound added. Moreover, cavitation is good for the dispersion of cyclic tetrameric units. The particle size of Zr-based colloidal sol can be tuned in the ranges of 1.5 to 120 nm by altering the molar ratio of precursor to precipitant, ultrasonic power density and radiation time. Meanwhile, ultrasonic power density and radiation time have effects on grain size and the crystalline transition temperature of particles which influence performance of the ceramic membrane. As a result, membranes exhibit high performance together with high permeability and desirable rejection. To develop such a simple and controllable method for tuning particle size is extremely important in the preparation of nanoporous ceramic membranes.

Direct Prediction of the Toxic Gas Diffusion Rule in a Real Environment Based on LSTM.

Predicting the diffusion rule of toxic gas plays a distinctly important role in emergency capability assessment and rescue work. Among diffusion prediction models, the traditional artificial neural network has exhibited excellent performance not only in prediction accuracy but also in calculation time. Nevertheless, with the continuous development of deep learning and data science, some new prediction models based on deep learning algorithms have been shown to be more advantageous because their structure can better discover internal laws and external connections between input data and output data. The long short-term memory (LSTM) network is a kind of deep learning neural network that has demonstrated outstanding achievements in many prediction fields. This paper applies the LSTM network directly to the prediction of toxic gas diffusion and uses the Project Prairie Grass dataset to conduct experiments. Compared with the Gaussian diffusion model, support vector machine (SVM) model, and back propagation (BP) network model, the LSTM model of deep learning has higher prediction accuracy (especially for the prediction at the point of high concentration values) while avoiding the occurrence of negative concentration values and overfitting problems found in traditional artificial neural network models.

A Context-Aware Edge-Based VANET Communication Scheme for ITS.

We propose a context-aware edge-based packet forwarding scheme for vehicular networks. The proposed scheme employs a fuzzy logic-based edge node selection protocol to find the best edge nodes in a decentralized manner, which can achieve an efficient use of wireless resources by conducting packet forwarding through edges. A reinforcement learning algorithm is used to optimize the last two-hop communications in order to improve the adaptiveness of the communication routes. The proposed scheme selects different edge nodes for different types of communications with different context information such as connection-dependency (connection-dependent or connection-independent), communication type (unicast or broadcast), and packet payload size. We launch extensive simulations to evaluate the proposed scheme by comparing with existing broadcast protocols and unicast protocols for various network conditions and traffic patterns.

Self-Cleaning Piezoelectric Membrane for Oil-in-Water Separation.

Ultrasound (US) treatment coupled with membrane filtration has been utilized for membrane fouling control in water treatment; however, large-scale implementation of ultrasonic cleaning equipment appeared to be cost-prohibitive. In this study, a porous lead zirconate titanate (PZT) membrane is presented that enables in situ ultrasound generation by the application of an alternating voltage (AV) to mitigate fouling during oil-in-water (O/W) emulsion separation. We expect that this method is much more cost-effective because it is more direct, avoiding buildup of fouling and the need to take the membrane offline. Because the PZT membrane is hydrophilic, its underwater surface is oleophobic so that the accumulated oil droplets will have little affinity and hence can be removed easily by in situ-generated US. The effect of the in situ US generation on membrane fouling was investigated through variation in the excitation AV and its frequency, O/W emulsion pH, emulsified oil concentration, crossflow velocity, and transmembrane pressure. The results indicated that the in situ US generation resulted in a substantial decrease of fouling during the filtration process of O/W emulsions, whereas the membrane flux was maintained closely at its initial value.

A reduced graphene oxide nanofiltration membrane intercalated by well-dispersed carbon nanotubes for drinking water purification.

In this study, we report a promising rGO-CNT hybrid nanofiltration (NF) membrane that was fabricated by loading reduced graphene oxide that was intercalated with carbon nanotubes (rGO-CNTs) onto an anodic aluminum oxide (AAO) microfiltration membrane via a facile vacuum-assisted filtration process. To create this NF membrane, the CNTs were first dispersed using block copolymers (BCPs); the effects of the types and contents of BCPs used on the dispersion of CNTs have been investigated. The as-prepared rGO-CNT hybrid NF membranes were then used for drinking water purification to retain the nanoparticles, dyes, proteins, organophosphates, sugars, and particularly humic acid. Experimentally, it is shown that the rGO-CNT hybrid NF membranes have high retention efficiency, good permeability and good anti-fouling properties. The retention was above 97.3% even for methyl orange (327 Da); for other objects, the retention was above 99%. The membrane's permeability was found to be as high as 20-30 L m(-2) h(-1) bar(-1). Based on these results, we can conclude that (i) the use of BCPs as a surfactant can enhance steric repulsion and thus disperse CNTs effectively; (ii) placing well-dispersed 1D CNTs within 2D graphene sheets allows an uniform network to form, which can provide many mass transfer channels through the continuous 3D nanostructure, resulting in the high permeability and separation performance of the rGO-CNT hybrid NF membranes.

Preparation and properties of Ni-doped ZnO rod arrays from aqueous solution.

Ni-doped ZnO rod arrays were successfully prepared on glass substrate from the aqueous solution at a temperature of 80 degrees C. The densities, diameters, and lengths of the rods can easily be well controlled through the concentrations of dopants, and the dopant Ni ions were incorporated into the wurtzite-structure of ZnO crystal. Room temperature photoluminescence spectrum of rod arrays show a strong emission band at 410 nm, and the oxygen deficiencies in ZnO structures were significantly reduced with Ni doped in ZnO rod arrays. The field dependence of magnetization measured at room temperature exhibited the obvious ferromagnetic properties.