Search

The electric field method proved in the lab and oil fields is an effective and fast way to significantly improve oil recovery, which can be applied to greatly realize the urgent-need requirements for energy, especially in tight sandstones. Generally, the changed molecular groups treated with an electric field modulate the wettability of reservoirs, affecting the final oil recovery. Herein, the investigation of the impact of the electric field on the molecular groups of reservoirs is imperative and meaningful. In this paper, tight sandstones were placed into a particular instrument and subjected to various strengths of the electric field. Nine treated powders and one untreated powder of tight sandstones were processed by Fourier transform infrared spectroscopy (FTIR) and X-ray photoelectron spectroscopy (XPS) experiments. FTIR results show that the electric field decreases aromatic groups, C-O groups, COOH groups, and aliphatic groups, whereas it increases C=C groups, C=O groups, and OH groups. Interestingly, the changes in C-O groups, C=O groups, COOH groups, and OH groups are all the competitive results of production and consumption during the treatment process. With regard to C-O groups and COOH groups, the consumption has an advantage over the production on the content of functional groups, and the situations for C==O groups and OH groups exhibit a contrary trend. The fitted result of XPS proves the fact that the electric field improves C=O groups, OH groups, and COOR groups, whereas it reduces C-O groups, supporting that the molecular groups can be mutually transformed during the electric field treatment. The obtained knowledge is beneficial to the study of electric field-related technologies on the molecular groups of reservoirs.

With the outbreak of COVID-19 in Wuhan, aggressive countermeasures have been taken, including the implementation of the unprecedented lockdown of the city, which will necessarily cause huge economic losses for the city of Wuhan. In this paper, we attempt to uncover the interactions between epidemic prevention and control measures and economic-social development by estimating the health loss and meso-economic loss from a human-oriented perspective. We implemented a compartmental model for the transmission dynamics and health burden assessment to evaluate the health losses, then estimated the direct and indirect economic losses of industries using the Input-Output model. Based on these estimates, the first monthly health losses and meso-economic losses caused by the lockdown was assessed. The overall policy effect of the lockdown policy in Wuhan was also investigated. The health loss and meso-economic losses are used to evaluate the health burden and loss of residents’ mental health, the direct economic loss of several worst-hit industries, and the indirect economic loss of all industries, respectively. Our findings reveal that the health burden caused by this pandemic is estimated to be 4.4899 billion yuan (CNY), and the loss of residents’ mental health is evaluated to be 114.545 billion yuan, the direct economic losses in transport, logistics, and warehousing, postal service, food, and beverage service industries reach 21.6094 billion yuan, and the monthly indirect economic losses of all industries are 36.39661994 billion yuan caused by the lockdown. The total monthly economic losses during the lockdown reach 177.0413 billion yuan. However, the lockdown policy has been considered to reduce COVID-19 infections by >180 thousand, which saves about 20 thousand lives, as well as nearly 30 billion yuan on medical costs. Therefore, the lockdown policy in Wuhan has obvious long-term benefits on the society and the total economic losses will be at a controllable level if effective measures are taken to combat COVID-19.

Capillary force and boundary layer effect are the main causes of non-Darcy flow in tight oil reservoir. This paper proposes a non-Darcy flow dynamics characterization method for low-speed water flooding in tight oil reservoirs. It applies constant-speed mercury injection and casting thin section experiments to quantitatively characterize the micro-pore throat structure parameters, and uses the visual experimental device to measure the boundary layer thickness and fit the expression of the relationship between boundary layer thickness and displacement pressure gradient and fluid viscosity. The results show that the ratio of boundary layer thickness to microtubule radius changes exponentially with the pressure gradient and fluid viscosity and that the boundary layer thickness decreases gradually with the increase of pressure gradient. Given the capillary force and boundary layer thickness, the flow rate of single capillary is calculated. On this basis, the equation of nonlinear seepage dynamic characteristics per unit area of core is derived by taking into account the throat distribution frequency and throat size characteristics. The new seepage flow model can reflect the nonlinear seepage flow law of tight oil reservoir and provide reference for parameter formulation during water flooding development of tight oil reservoir.

Understanding and manipulating wettability alterations has tremendous implications in theoretical research and industrial applications. This study proposes a novel idea of applying ultrasonic for wettability alterations and also provides its quantitative characterizations and in-depth analyses. More specifically, with pretreatment of ultrasonic, mechanisms of wettability alteration were characterized from the contact angle measurements, as well as the in-depth analyses from atomic force microscopy (AFM), X-ray diffraction (XRD), and Fourier-transform infrared spectroscopy (FTIR). After ultrasonic treatments, the wettability of mineral with low permeability is determined to altered from strong hydrophilic to intermediate wettability. The mechanism interpretations are conducted by means of the AFM, XRD, and FTIR. Basically, as the time of ultrasonic treatment increases, the AFM results indicate that the roughness of rock surface and oil/rock interface (contact area) with surroundings of brine is enhanced. Meanwhile, the XRD results show the diffusions of clays from the rock surface to the aqueous phase, and FTIR indicates that the number of functional groups of Si-O-Si, C-O-C, C-O, C═O, and OH decreases while the number of COOH and C═C═O groups increases. This study clearly reveals the surface chemistry of oil-rock wettability alteration in the subsurface conditions, which would provide technical support for subsurface usage of geo-energy productions and carbon sequestrations.

Existing road pothole detection approaches can be classified as computer vision-based or machine learning-based. The former approaches typically employ 2-D image analysis/understanding or 3-D point cloud modeling and segmentation algorithms to detect road potholes from vision sensor data. The latter approaches generally address road pothole detection using convolutional neural networks (CNNs) in an end-to-end manner. However, road potholes are not necessarily ubiquitous and it is challenging to prepare a large well-annotated dataset for CNN training. In this regard, while computer vision-based methods were the mainstream research trend in the past decade, machine learning-based methods were merely discussed. Recently, we published the first stereo vision-based road pothole detection dataset and a novel disparity transformation algorithm, whereby the damaged and undamaged road areas can be highly distinguished. However, there are no benchmarks currently available for state-of-the-art (SoTA) CNNs trained using either disparity images or transformed disparity images. Therefore, in this paper, we first discuss the SoTA CNNs designed for semantic segmentation and evaluate their performance for road pothole detection with extensive experiments. Additionally, inspired by graph neural network (GNN), we propose a novel CNN layer, referred to as graph attention layer (GAL), which can be easily deployed in any existing CNN to optimize image feature representations for semantic segmentation. Our experiments compare GAL-DeepLabv3+, our best-performing implementation, with nine SoTA CNNs on three modalities of training data: RGB images, disparity images, and transformed disparity images. The experimental results suggest that our proposed GAL-DeepLabv3+ achieves the best overall pothole detection accuracy on all training data modalities., accepted as a regular paper to IEEE Transactions on Image Processing

Through the platform of smart education, teachers and students can enjoy equal access to knowledge. This is helpful to solve the problem of information asymmetry between teachers and students in teaching. This new education and teaching system can improve students' participation in curriculum learning and cultivate their hard work spirit in professional learning. In order to recognize and classify actions accurately, this paper chooses the EM algorithm based on Gaussian mixture model which is an iterative optimization algorithm. Each iteration is divided into expectation step and maximum likelihood step. Finally, the optimal convergence value can be found according to each iteration step.

CO2 enhanced oil recovery (EOR) has been proven its capability to explore the unconventional tight oil reservoirs and potential for geological carbon storage. Meanwhile, the extremely low permeability pores exaggerate the difficulty CO2 EOR and geological storage processing in the actual field. This paper initiates the ultrasonic-assisted approach to facilitate the oil-gas miscibility development and finally contribute to unlock more tight oils. First, the physical properties of crude oil with and without ultrasonic treatments were experimentally analysed through gas chromatography (GC), Fourier-transform infrared spectroscopy (FTIR) and viscometer. Second, the oil-gas minimum miscibility pressures (MMPs) were measured from the slim-tube test and the miscibility developments with and without ultrasonic treatments were interpreted from the mixing-cell method. Third, the nuclear-magnetic resonance (NMR) assisted coreflood tests were conducted to physically model the recovery process in porous media and directly obtain the recovery factor. Basically, the ultrasonic treatment (40KHz and 200W for 8 hours) was found to substantially change the oil properties, with viscosity (at 60°C) reduced from 4.1 to 2.8mPa·s, contents of resin and asphaltene decreased from 27.94% and 6.03% to 14.2% and 3.79%, respectively. The FTIR spectrum shows the unsaturated C-H bond, C-O bond and C≡C bond in macromolecules were broken from ultrasonic, which caused the macromolecules (e.g., resin and asphaltenes) to be decomposed into smaller carbon-number molecules. Accordingly, the MMP was determined to be reduced from 15.8 to 14.9MPa from the slim-tube test and the oil recovery factor increased by over 10%. This study reveals the mechanisms of ultrasonic-assisted CO2 miscible EOR in producing tight oils.

Convolutional neural network (CNN)-based stereo matching approaches generally require a dense cost volume (DCV) for disparity estimation. However, generating such cost volumes is computationally-intensive and memory-consuming, hindering CNN training and inference efficiency. To address this problem, we propose SCV-Stereo, a novel CNN architecture, capable of learning dense stereo matching from sparse cost volume (SCV) representations. Our inspiration is derived from the fact that DCV representations are somewhat redundant and can be replaced with SCV representations. Benefiting from these SCV representations, our SCV-Stereo can update disparity estimations in an iterative fashion for accurate and efficient stereo matching. Extensive experiments carried out on the KITTI Stereo benchmarks demonstrate that our SCV-Stereo can significantly minimize the trade-off between accuracy and efficiency for stereo matching. Our project page is https://sites.google.com/view/scv-stereo., 5 pages, 3 figures and 2 tables. This paper is accepted by ICIP 2021

Recently, deep-learning based approaches have achieved impressive performance for autonomous driving. However, end-to-end vision-based methods typically have limited interpretability, making the behaviors of the deep networks difficult to explain. Hence, their potential applications could be limited in practice. To address this problem, we propose an interpretable end-to-end vision-based motion planning approach for autonomous driving, referred to as IVMP. Given a set of past surrounding-view images, our IVMP first predicts future egocentric semantic maps in bird's-eye-view space, which are then employed to plan trajectories for self-driving vehicles. The predicted future semantic maps not only provide useful interpretable information, but also allow our motion planning module to handle objects with low probability, thus improving the safety of autonomous driving. Moreover, we also develop an optical flow distillation paradigm, which can effectively enhance the network while still maintaining its real-time performance. Extensive experiments on the nuScenes dataset and closed-loop simulation show that our IVMP significantly outperforms the state-of-the-art approaches in imitating human drivers with a much higher success rate. Our project page is available at https://sites.google.com/view/ivmp., 7 pages, 5 figures and 1 table. This paper is accepted by ICRA 2021. arXiv admin note: text overlap with arXiv:2104.08862

As an effective risk management mechanism, agricultural insurance can reduce the risk of uncertainty in agricultural production and guarantee food security. Based on Chinese provincial panel data from 2003 to 2020, this study uses the Entropy Method to measure food security and systematically examines the impact of climate change and agricultural insurance on food security as well as its mechanisms. The present study found that climate change, especially extreme temperatures, has a significant negative impact on food security and food production. The promotion effect of agricultural insurance on food security increases with increased investments in technology, education, and other factors. Furthermore, our findings suggest the presence of geographical variations in the contribution of agricultural insurance to ensuring food security, with greater coverage in major food-producing regions. Additionally, maize yields are better protected by agricultural insurance than wheat and rice yields. To encourage sustainable agricultural development, the Chinese government should set up a diversified subsidy scheme with various planting scales and plant structures.

This study examines the spatial and temporal impacts of climate change on grain production in China. This is achieved by establishing a spatial error model consisting of four indicators: the climate, air pollution, economic behavior, and agricultural technology, covering 31 provinces in China from 2004 to 2020. These indicators are used to validate the spatial impacts of climate change on grain production. Air pollution data are used as instrumental variables to address the causality between climate and grain production. The regression results show that: First, climatic variables all have a non-linear “increasing then decreasing” effect on food production. Second, SO2, PM10, and PM2.5 have a negative impact on grain production. Based on the model, changes in the climatic production potential of grain crops can be calculated, and the future spatial layout of climate production can also be predicted by using random forests. Studies have shown that the median value of China’s grain production potential is decreasing, and the low value is increasing.

This paper proposes three-filters-to-normal (3F2N), an accurate and ultrafast surface normal estimator (SNE), which is designed for structured range sensor data, e.g., depth/disparity images. 3F2N SNE computes surface normals by simply performing three filtering operations (two image gradient filters in horizontal and vertical directions, respectively, and a mean/median filter) on an inverse depth image or a disparity image. Despite the simplicity of 3F2N SNE, no similar method already exists in the literature. To evaluate the performance of our proposed SNE, we created three large-scale synthetic datasets (easy, medium and hard) using 24 3D mesh models, each of which is used to generate 1800--2500 pairs of depth images (resolution: 480X640 pixels) and the corresponding ground-truth surface normal maps from different views. 3F2N SNE demonstrates the state-of-the-art performance, outperforming all other existing geometry-based SNEs, where the average angular errors with respect to the easy, medium and hard datasets are 1.66 degrees, 5.69 degrees and 15.31 degrees, respectively. Furthermore, our C++ and CUDA implementations achieve a processing speed of over 260 Hz and 21 kHz, respectively. Our datasets and source code are publicly available at sites.google.com/view/3f2n., Comment: webpage: sites.google.com/view/3f2n, accepted to IEEE RA-L and ICRA'21

The offloading of computing tasks in edge computing has always been a research hotspot and difficulty in recent years. As an effective way to run various applications on mobile devices with limited resources, it has been extensively studied by scholars from all walks of life. However, the traditional ground-based network-based edge computing network architecture cannot meet the needs of edge users with limited geographic areas. Therefore, this paper proposes an LEO (low earth orbit) satellite-based multiaccess edge computing network architecture and establishes a differential game model for this architecture. To obtain the Nash equilibrium solution of the open loop and the Nash equilibrium solution of the feedback for the task offloading amount, the relationship between the user’s income and the QoE level under the optimal task offloading amount is finally analyzed and discussed.

Path planning is a fundamental capability for autonomous navigation of robotic wheelchairs. With the impressive development of deep-learning technologies, imitation learning-based path planning approaches have achieved effective results in recent years. However, the disadvantages of these approaches are twofold: 1) they may need extensive time and labor to record expert demonstrations as training data; and 2) existing approaches could only receive high-level commands, such as turning left/right. These commands could be less sufficient for the navigation of mobile robots (e.g., robotic wheelchairs), which usually require exact poses of goals. We contribute a solution to this problem by proposing S2P2, a self-supervised goal-directed path planning approach. Specifically, we develop a pipeline to automatically generate planned path labels given as input RGB-D images and poses of goals. Then, we present a best-fit regression plane loss to train our data-driven path planning model based on the generated labels. Our S2P2 does not need pre-built maps, but it can be integrated into existing map-based navigation systems through our framework. Experimental results show that our S2P2 outperforms traditional path planning algorithms, and increases the robustness of existing map-based navigation systems. Our project page is available at https://sites.google.com/view/s2p2., Comment: 7 pages, 6 figures and 3 tables. This paper is accepted by ICRA 2021

Supervised learning with deep convolutional neural networks (DCNNs) has seen huge adoption in stereo matching. However, the acquisition of large-scale datasets with well-labeled ground truth is cumbersome and labor-intensive, making supervised learning-based approaches often hard to implement in practice. To overcome this drawback, we propose a robust and effective self-supervised stereo matching approach, consisting of a pyramid voting module (PVM) and a novel DCNN architecture, referred to as OptStereo. Specifically, our OptStereo first builds multi-scale cost volumes, and then adopts a recurrent unit to iteratively update disparity estimations at high resolution; while our PVM can generate reliable semi-dense disparity images, which can be employed to supervise OptStereo training. Furthermore, we publish the HKUST-Drive dataset, a large-scale synthetic stereo dataset, collected under different illumination and weather conditions for research purposes. Extensive experimental results demonstrate the effectiveness and efficiency of our self-supervised stereo matching approach on the KITTI Stereo benchmarks and our HKUST-Drive dataset. PVStereo, our best-performing implementation, greatly outperforms all other state-of-the-art self-supervised stereo matching approaches. Our project page is available at sites.google.com/view/pvstereo., Comment: 8 pages, 8 figures and 2 tables. This paper is accepted by IEEE RA-L with ICRA 2021

Traditional decision and planning frameworks for self-driving vehicles (SDVs) scale poorly in new scenarios, thus they require tedious hand-tuning of rules and parameters to maintain acceptable performance in all foreseeable cases. Recently, self-driving methods based on deep learning have shown promising results with better generalization capability but less hand engineering effort. However, most of the previous learning-based methods are trained and evaluated in limited driving scenarios with scattered tasks, such as lane-following, autonomous braking, and conditional driving. In this paper, we propose a graph-based deep network to achieve scalable self-driving that can handle massive traffic scenarios. Specifically, more than 7,000 km of evaluation is conducted in a high-fidelity driving simulator, in which our method can obey the traffic rules and safely navigate the vehicle in a large variety of urban, rural, and highway environments, including unprotected left turns, narrow roads, roundabouts, and pedestrian-rich intersections. Demonstration videos are available at https://caipeide.github.io/dignet/., IROS 2021, 6 pages

Freespace detection is an essential component of visual perception for self-driving cars. The recent efforts made in data-fusion convolutional neural networks (CNNs) have significantly improved semantic driving scene segmentation. Freespace can be hypothesized as a ground plane, on which the points have similar surface normals. Hence, in this paper, we first introduce a novel module, named surface normal estimator (SNE), which can infer surface normal information from dense depth/disparity images with high accuracy and efficiency. Furthermore, we propose a data-fusion CNN architecture, referred to as RoadSeg, which can extract and fuse features from both RGB images and the inferred surface normal information for accurate freespace detection. For research purposes, we publish a large-scale synthetic freespace detection dataset, named Ready-to-Drive (R2D) road dataset, collected under different illumination and weather conditions. The experimental results demonstrate that our proposed SNE module can benefit all the state-of-the-art CNNs for freespace detection, and our SNE-RoadSeg achieves the best overall performance among different datasets., ECCV 2020

In this paper, we introduce a novel suspect-and-investigate framework, which can be easily embedded in a drone for automated parking violation detection (PVD). Our proposed framework consists of: 1) SwiftFlow, an efficient and accurate convolutional neural network (CNN) for unsupervised optical flow estimation; 2) Flow-RCNN, a flow-guided CNN for car detection and classification; and 3) an illegally parked car (IPC) candidate investigation module developed based on visual SLAM. The proposed framework was successfully embedded in a drone from ATG Robotics. The experimental results demonstrate that, firstly, our proposed SwiftFlow outperforms all other state-of-the-art unsupervised optical flow estimation approaches in terms of both speed and accuracy; secondly, IPC candidates can be effectively and efficiently detected by our proposed Flow-RCNN, with a better performance than our baseline network, Faster-RCNN; finally, the actual IPCs can be successfully verified by our investigation module after drone re-localization., 17 pages, 11 figures and 3 tables. This paper is accepted by ECCV Workshops 2020

Manual visual inspection performed by certified inspectors is still the main form of road pothole detection. This process is, however, not only tedious, time-consuming and costly, but also dangerous for the inspectors. Furthermore, the road pothole detection results are always subjective, because they depend entirely on the individual experience. Our recently introduced disparity (or inverse depth) transformation algorithm allows better discrimination between damaged and undamaged road areas, and it can be easily deployed to any semantic segmentation network for better road pothole detection results. To boost the performance, we propose a novel attention aggregation (AA) framework, which takes the advantages of different types of attention modules. In addition, we develop an effective training set augmentation technique based on adversarial domain adaptation, where the synthetic road RGB images and transformed road disparity (or inverse depth) images are generated to enhance the training of semantic segmentation networks. The experimental results demonstrate that, firstly, the transformed disparity (or inverse depth) images become more informative; secondly, AA-UNet and AA-RTFNet, our best performing implementations, respectively outperform all other state-of-the-art single-modal and data-fusion networks for road pothole detection; and finally, the training set augmentation technique based on adversarial domain adaptation not only improves the accuracy of the state-of-the-art semantic segmentation networks, but also accelerates their convergence., Comment: 16 pages, 7 figures and 2 tables. This paper is accepted by ECCV Workshops 2020

The segmentation of drivable areas and road anomalies are critical capabilities to achieve autonomous navigation for robotic wheelchairs. The recent progress of semantic segmentation using deep learning techniques has presented effective results. However, the acquisition of large-scale datasets with hand-labeled ground truth is time-consuming and labor-intensive, making the deep learning-based methods often hard to implement in practice. We contribute to the solution of this problem for the task of drivable area and road anomaly segmentation by proposing a self-supervised learning approach. We develop a pipeline that can automatically generate segmentation labels for drivable areas and road anomalies. Then, we train RGB-D data-based semantic segmentation neural networks and get predicted labels. Experimental results show that our proposed automatic labeling pipeline achieves an impressive speed-up compared to manual labeling. In addition, our proposed self-supervised approach exhibits more robust and accurate results than the state-of-the-art traditional algorithms as well as the state-of-the-art self-supervised algorithms., Published in IEEE Robotics and Automation Letters (RA-L); 8 pages, 8 figures and 3 tables

Traditional methods for autonomous driving are implemented with many building blocks from perception, planning and control, making them difficult to generalize to varied scenarios due to complex assumptions and interdependencies. Recently, the end-to-end driving method has emerged, which performs well and generalizes to new environments by directly learning from export-provided data. However, many existing methods on this topic neglect to check the confidence of the driving actions and the ability to recover from driving mistakes. In this paper, we develop an uncertainty-aware end-to-end trajectory generation method based on imitation learning. It can extract spatiotemporal features from the front-view camera images for scene understanding, and then generate collision-free trajectories several seconds into the future. The experimental results suggest that under various weather and lighting conditions, our network can reliably generate trajectories in different urban environments, such as turning at intersections and slowing down for collision avoidance. Furthermore, closed-loop driving tests suggest that the proposed method achieves better cross-scene/platform driving results than the state-of-the-art (SOTA) end-to-end control method, where our model can recover from off-center and off-orientation errors and capture 80% of dangerous cases with high uncertainty estimations., 11 pages, 14 figures, and 4 tables. The paper is accepted by IEEE Transactions on Intelligent Vehicles (T-IV), 2020

Atomic force microscopy (AFM), which allows 3D construction, nondestructive, and has a high resolution has been used to describe the pore structure of unconventional reservoirs. Due to its scalability, convenience, and flexibility, scanning electron microscopy (SEM) is widely conducted to describe the surface morphology of unconventional reservoirs. However, the two technologies are rarely combined to achieve fast and effective observations of the pore structure of unconventional reservoirs at the same location of the same sample (SLSS). In this study, nanoindentation technology is introduced to create a marked district (MD). Afterward, fast and effective observations of the pore structure of tight sandstones at the SLSS can be realized by searching this identifiable district with AFM and cold-field scanning electron microscopy (CF-SEM). The results demonstrate that AFM and CF-SEM rapidly capture observations of the surface morphology of tight sandstones at the SLSS. The method used in this study accomplishes the goal of fast and effective scanning at the SLSS (only a few minutes), while exhibiting the high potential of combining AFM and CF-SEM for comprehensively investigating the pore structure of reservoirs ranging from nanoscale to microscale. The most unique component of the presented method is that it is especially for the in situ observation of variations in the surface properties (morphology, electric, mechanical, surface force, and roughness) and pore size of all heterogeneous materials subjected to various external stimuli because the applied experimental technologies (nanoindentation, AFM, and CF-SEM) are all exhibit nondestructive properties.

Carbon dioxide miscible flooding has been proven to be one of the most effective enhanced oil recovery (EOR) technologies, particularly for light and medium oil reservoirs. However, specific effects of pore structure on CO2 miscible flooding recovery in low permeability reservoir lack in-depth understandings. In this paper, pore structures are specifically studied by means of the molecular mechanics to evaluate their effects on the CO2 EOR in the low permeability reservoir. First, a series lab experiments are performed for the pore and fluid characterization. More specifically, the pore throat size and distribution frequency are measured from the high-pressure mercury injection and nuclear magnetic resonance. The minimum miscibility pressure is determined from the slim-tube tests with known oil compositions tested from gas chromatography analysis. Second, the regularity of CO2 extraction is explored on the basis of molecular mechanics and the thickness of raffinate is calculated. Finally, the raffinate volume and recovery ratio in the pores are calculated after the CO2 miscible flooding. The results show that a raffinate-layer with thickness of 0.13 μm remains on the surface of the pore after the CO2 miscible flooding, which would cause the oil to be immobile since the throat could be blocked when the throat radius is smaller than 0.26 μm. The recoveries of cores C-1 and C-2 are 70.1 % and 61.4 % from calculations and 68.4 % and 59.8 % from experiments, whose errors are 2.5 % and 2.7 %, respectively. This study would be beneficial to analyze the CO2 miscible flooding in reservoirs with different pore structures and provide technical support for improving CO2 utilization efficiency.

Collision-free space detection is a critical component of autonomous vehicle perception. The state-of-the-art algorithms are typically based on supervised learning. The performance of such approaches is always dependent on the quality and amount of labeled training data. Additionally, it remains an open challenge to train deep convolutional neural networks (DCNNs) using only a small quantity of training samples. Therefore, this paper mainly explores an effective training data augmentation approach that can be employed to improve the overall DCNN performance, when additional images captured from different views are available. Due to the fact that the pixels of the collision-free space (generally regarded as a planar surface) between two images captured from different views can be associated by a homography matrix, the scenario of the target image can be transformed into the reference view. This provides a simple but effective way of generating training data from additional multi-view images. Extensive experimental results, conducted with six state-of-the-art semantic segmentation DCNNs on three datasets, demonstrate the effectiveness of our proposed training data augmentation algorithm for enhancing collision-free space detection performance. When validated on the KITTI road benchmark, our approach provides the best results for stereo vision-based collision-free space detection., Comment: accepted to IEEE/ASME Transactions on Mechatronics

Since early 2020, the coronavirus disease 2019 (COVID-19) has spread rapidly across the world. As at the date of writing this article, the disease has been globally reported in 223 countries and regions, infected over 108 million people and caused over 2.4 million deaths (https://covid19.who.int/, accessed on Feb. 17, 2021). Avoiding person-to-person transmission is an effective approach to control and prevent the pandemic. However, many daily activities, such as transporting goods in our daily life, inevitably involve person-to-person contact. Using an autonomous logistic vehicle to achieve contact-less goods transportation could alleviate this issue. For example, it can reduce the risk of virus transmission between the driver and customers. Moreover, many countries have imposed tough lockdown measures to reduce the virus transmission (e.g., retail, catering) during the pandemic, which causes inconveniences for human daily life. Autonomous vehicle can deliver the goods bought by humans, so that humans can get the goods without going out. These demands motivate us to develop an autonomous vehicle, named as Hercules, for contact-less goods transportation during the COVID-19 pandemic. The vehicle is evaluated through real-world delivering tasks under various traffic conditions.

During the extraction process of LX high and low pressure reservoirs, gas backflow happens, resulting in the loss of capacity; meanwhile, low-pressure formations will be harmed. The method, which combines the experimental study and theoretical analysis, determines rational proration of research area and directs the efficient exploitation of tight gas reservoir with producing water. Establishing the productivity formula of multilayer reservoir with producing water determines the upper limitation of production allocation of water-producing gas well; moreover, according to the IPR curve, the lower limitation of rational proration of water-producing gas well has determined. Finally, the result indicates that the rational proration of the type I well in research area is 3.0–3.2 × 104 m3 and the rational proration of the type II well in research area is 1.6–2.0 × 104 m3. The case study results indicate that the production allocation of multilayer reservoir with producing water can effectively control the water production rate and Wellbore hydrops, reduce the effect of gas backflow, as well as maintain the smooth production of gas wells if the production allocation which was calculated by the new methods is within the reasonable capacity range.

The joint detection of drivable areas and road anomalies is a crucial task for ground mobile robots. In recent years, many impressive semantic segmentation networks, which can be used for pixel-level drivable area and road anomaly detection, have been developed. However, the detection accuracy still needs improvement. Therefore, we develop a novel module named the Normal Inference Module (NIM), which can generate surface normal information from dense depth images with high accuracy and efficiency. Our NIM can be deployed in existing convolutional neural networks (CNNs) to refine the segmentation performance. To evaluate the effectiveness and robustness of our NIM, we embed it in twelve state-of-the-art CNNs. The experimental results illustrate that our NIM can greatly improve the performance of the CNNs for drivable area and road anomaly detection. Furthermore, our proposed NIM-RTFNet ranks 8th on the KITTI road benchmark and exhibits a real-time inference speed., Comment: 6 pages, 6 figures and 1 table. This paper is accepted by IROS 2020

The interpretation of ego motion and scene change is a fundamental task for mobile robots. Optical flow information can be employed to estimate motion in the surroundings. Recently, unsupervised optical flow estimation has become a research hotspot. However, unsupervised approaches are often easy to be unreliable on partially occluded or texture-less regions. To deal with this problem, we propose CoT-AMFlow in this paper, an unsupervised optical flow estimation approach. In terms of the network architecture, we develop an adaptive modulation network that employs two novel module types, flow modulation modules (FMMs) and cost volume modulation modules (CMMs), to remove outliers in challenging regions. As for the training paradigm, we adopt a co-teaching strategy, where two networks simultaneously teach each other about challenging regions to further improve accuracy. Experimental results on the MPI Sintel, KITTI Flow and Middlebury Flow benchmarks demonstrate that our CoT-AMFlow outperforms all other state-of-the-art unsupervised approaches, while still running in real time. Our project page is available at https://sites.google.com/view/cot-amflow., Comment: 13 pages, 3 figures and 6 tables. This paper is accepted by CoRL 2020

The size of pore throats in tight oil reservoirs varies from a few tens of nanometers to several hundred micrometers. This complex and diverse microscopic pore throat structure restricts the exploration and development process, affecting the recovery rate. Size, shape, and spatial distribution of pore throats in tight oil reservoirs are revealed in this paper via a scanning electron microscope (SEM), casting thin-sections, high-pressure mercury injection, and Nano-CT. Results show that the pore type of the Chang 7 tight oil reservoir in the Xin’anbian area is mainly divided into three categories: intergranular pores, dissolution pores, and microfractures. Numerous nanoscale pore throats have developed in this area, which greatly contributes to reservoir capacity and seepage. Pore throat distribution on capillary pressure curves of different samples shows that when the threshold pressure is less than 1 Mpa, there are many micron-sized pores with good connectivity and pore throats form a large tubular shape with a throat radius between 3.6 and 1064 nm. When the threshold pressure ranges from 1 to 3 MPa, there are many nanoscale pores with good local connectivity, intragranular dissolution pores develop, and pore throats are in tube bundles and spherical in shape with the throat radius between 3.6−657 nm. When the threshold pressure is greater than 3 MPa, nanoscale microfractures develop and are connected to neighboring small spherical pores, small spherical nanopore spaces become isolated, vertical connectivity is poor, and the throat radius is between 3.6−242 nm.

CO2 enhanced oil recovery (EOR) has proven its capability to explore unconventional tight oil reservoirs and the potential for geological carbon storage. Meanwhile, the extremely low permeability pores increase the difficulty of CO2 EOR and geological storage processing in the actual field. This paper initiates the ultrasonic-assisted approach to facilitate oil–gas miscibility development and finally contributes to excavating more tight oils. Firstly, the physical properties of crude oil with and without ultrasonic treatments were experimentally analyzed through gas chromatography (GC), Fourier-transform infrared spectroscopy (FTIR) and viscometer. Secondly, the oil–gas minimum miscibility pressures (MMPs) were measured from the slim-tube test and the miscibility developments with and without ultrasonic treatments were interpreted from the mixing-cell method. Thirdly, the nuclear-magnetic resonance (NMR) assisted coreflood tests were conducted to physically model the recovery process in porous media and directly obtain the recovery factor. Basically, the ultrasonic treatment (40 KHz and 200 W for 8 h) was found to substantially change the oil properties, with viscosity (at 60 °C) reduced from 4.1 to 2.8 mPa·s, contents of resin and asphaltene decreased from 27.94% and 6.03% to 14.2% and 3.79%, respectively. The FTIR spectrum showed that the unsaturated C-H bond, C-O bond and C≡C bond in macromolecules were broken from the ultrasonic, which caused the macromolecules (e.g., resin and asphaltenes) to be decomposed into smaller carbon-number molecules. Accordingly, the MMP was determined to be reduced from 15.8 to 14.9 MPa from the slim-tube test and the oil recovery factor increased by an additional 11.7%. This study reveals the mechanisms of ultrasonic-assisted CO2 miscible EOR in producing tight oils.

This paper proposes a novel reconfigurable ankle rehabilitation robot based on parallel mechanism. The proposed robot is composed of the linear actuators, reconfigurable mechanism, connecting rods, rockers and the moving platform. Then, the kinematic characteristics analysis for the proposed robot is given. The function relationships between the moving platform and the actuators are obtained. Furthermore, the verification for this novel reconfigurable robot is completed by wireless movement capture instrument, and the results show that the prototype ankle rehabilitation robot can meet the motion needed for the ankle rehabilitation.

This paper investigates the characteristics of parallel connected three phase inverters based on modular design and distributed control. Every inverter module has a local hardware manager and the whole system has a central application manager. The hardware managers and application manager are connected together by the optical fiber to build a fiber ring net structure. The fiber ring net can achieve high communication rate and synchronous drive signals in different local controllers. Application manager will decide the number of operating modules and every inverter module can run or stop dynamically. The system structure and system model vary with the number of operating modules. The controller in application manager should be designed based on every possible model. In other words, the controller should adapt to the worst situation. In this converter, current sharing performance is good with no current sharing method used. A prototype is made to verify the design and analysis.

A bidirectional converter for electromechanical brake/the recycling of power generation energy is proposed in this paper. The gain of the converter is high. Since the converter use the technique of passive lossless buffer and synchronous rectification, the main switches achieve the technique of soft switching. Meanwhile the conversation efficiency is high too. Some experimental results and analysis of theory are given to verify the effectiveness of the proposed bidirectional converter.

Based on boost converter operating in discontinuous mode, this paper proposes an energy balance model (EBM) for analyzing bifurcation and chaos phenomena of capacitor energy and output voltage when the converter parameter is varying. It is found that the capacitor energy and output voltage dynamic behaviors exhibit the typical period- doubling route to chaos by increasing the feedback gain constant K of proportional controller. The accurate position of the first bifurcation point and the iterative diagram of the capacitor energy with every K can be derived from EBM. Finally, the underlying causes for bifurcations and chaos of a general class of nonlinear systems such as power converters are analyzed from the energy balance viewpoint. Comparing with the discrete iterative model, EBM is simple and high accuracy .This model can be easily developed on the nonlinear study of the other converters.

A theoretical study was conducted using a multiple-voltage probe sensor for detecting nonconducting inclusions in conducting media. Results show that the multiple-voltage probe sensor is capable of providing precise quantitative measurements of submerged nonconducting objects if the surface voltage response has a standard two-peak form. The standard response is observed for well-localized non-slender single inclusions below the sensor surface. In this case, the peak separation distance is associated with the inclusion depth whereas the peak magnitude is associated with the inclusion volume. Linear dependencies of the inclusion depth and the inclusion volume are observed for a wide variety of inclusion shapes. The predefined form of the surface voltage response makes it feasible to identify useful signal responses at very high noise levels. This is accomplished by using a 2D neural network classifier, based on the probabilistic neural network. A reasonable recognition error of less than 20 % is obtained if the signal-to-noise ratio is larger than or equal to 1/10. A metal casting example shows that the multiple-voltage probe sensor can measure inclusion concentrations in hot conducting melts (gas bubbles and sludge) with inclusion radii in the range from 100 to 1000 μm. In contrast to existing particle counter technology, this sensor construction is simple to construct and does not require special aperture and vacuum treatment.