Search

Automatic driving technology has high accuracy and real-time requirements for pedestrian identification and localization. Pedestrian detection is a basic and necessary function in vision-based pedestrian detection systems and collision warning, which can effectively avoid traffic accidents and improve road driving safety to a certain extent. In this paper, a lightweight solution based on the FCOS-DEFPN model is proposed for real-time pedestrian detection. Based on the FCOS model, this paper proposes the FCOS-DEFPN model, which achieves the lightweight of the network by replacing the ResNet50 backbone network with the MobilenetV3 network and using the depth separable convolution instead of the ordinary convolution for parameter compression. While maintaining the detection accuracy, this paper introduces data enhancement methods such as Random Erasing and Morsia to simulate pedestrian occlusion and small target scenarios to improve the robustness of the model. For the pedestrian occlusion scenario, this paper introduces a lightweight attention network ECA, which helps to extract pedestrian features better. For small-target multi-scale pedestrians, the DEFPN feature pyramid network is proposed, which acquires feature information at multiple scales by attentional fusion of feature layers at different scales from top-down, bottom-up, and front-back. The experimental results show that the proposed model is enhanced in terms of detection accuracy for occluded and small-target pedestrians, and satisfies real-time pedestrian detection under the premise of robustness in complex scenes.

Pedestrian trajectory prediction is a key technology in surveillance systems and autonomous driving. However, due to the high uncertainty and dynamic spatial-temporal dependence of pedestrian movement, timely and accurate pedestrian trajectory prediction, especially long-term prediction, is still an open challenge. However, the existing models lack an effective modeling method for temporal dependence and spatial interaction modeling. To solve these problems, in this paper, we propose a novel paradigm of Sparse Transformer Networks with Spatial-Temporal Graph (STGSTN), which captures complex spatial-temporal interactions by stacking spatial-temporal Transformer blocks and improves the accuracy of pedestrian trajectory prediction by combining dynamic spatial dependence and long-range temporal dependence. We propose a new variant of sparse spatial transformer combined with graph neural networks, which uses the self-attention mechanism to dynamically model spatial dependencies to capture the state of pedestrian movement. The multi-head attention mechanism is also used to jointly model various patterns of spatial dependence. In addition, the sparse temporal transformer is used to learn the sparse attention map for time modeling, and then the long-range temporal dependence between pedestrians is modeled. Compared with the existing work, STGSTN can effectively and efficiently train long-range spatial-temporal dependencies. Experimental results show that STGSTN is competitive with state-of-the-art techniques on the ETH-UCY dataset, especially for long-term prediction. The performance of STGSTN may be limited by the quality of the training data, and its generalization to highly dynamic or crowded environments remains a challenge. Future work will focus on addressing these limitations and exploring additional contextual information to further enhance prediction accuracy.

Millions of lives are lost in road accidents annually, underscoring the severity of traffic incidents. Despite numerous studies on crash severity forecasting with static ensemble machine learning, the potential of Dynamic Ensemble Selection (DES) for combining classifiers and improving performance remains relatively unexplored in this field. In contrast to previous research focused on static ensemble machine learning models, our study introduces an interpretable Bayesian Optimized Dynamic Ensemble Selection strategy, reinforced with instance hardness and region of competence considerations, for crash risk assessment. Utilizing six algorithms – Random Forest, Extra Tree, Cat Boost, Light GBM, KT Boost, and XGBoost – optimized through cross-validation, we employed bagging as a hybrid ensemble selection criterion to ensure ensemble diversity for DES models. Addressing class imbalance with cost-sensitive learning, we evaluated six DES algorithms, ‘KNORA-E, KNORA-U, DES-KNN, DES-RRC, DES-KL, and META-DES’ using a pool of input classifiers, based on key metrics such as Balanced Accuracy Score, Sensitivity, G-mean, and Matthews Correlation Coefficient. Our results revealed that the KNORA-E ensemble with Cat Boost achieved the highest performance, with a Balanced Accuracy (BCA) of 69.3%, Matthews Correlation Coefficient (MCC) of 0.36, and a G-mean score of 69.10%. DES-RRC demonstrated notable performance across all base classifiers, particularly in addressing imbalanced classes, and DES-RRC with LightGBM emerged as the second-best performer. SHAP analysis for the KNORAE-Cat Boost model identified key factors in crash severity as, road user gender, airbag deployment, vehicle age and road surface conditions. This approach could be a valuable tool for informed decision-making and targeted interventions.

In this research, we present an advanced predictive framework designed to assess the turbulence induced by low-level wind shear near the runways at Hong Kong International Airport (HKIA), utilizing data from Pilot Reports (PIREP). This framework integrates the TabNet architecture with SHapley Additive exPlanations (SHAP), thereby enhancing both predictive accuracy and interpretability. Given the imbalance in the PIREP data, we implement various data augmentation techniques and employ Bayesian optimization to fine-tune the hyperparameters of the TabNet model. Our analysis revealed that the TabNet model, when applied to data balanced with the Support Vector Machine - Synthetic Minority Over-sampling Technique (SVM-SMOTE), demonstrated superior performance, evidenced by a Geometric Mean (G-Mean) of 0.74, a Matthews Correlation Coefficient (MCC) of 0.37, a Balanced Accuracy (BA) of 0.74, and an Area Under the Receiver Operating Characteristic (AU-ROC) of 0.739. SHAP analysis further enhanced the interpretability of the TabNet model by identifying key contributing factors, including the magnitude of wind shear and altitude, which significantly influence the likelihood of significant turbulence occurrence. Specifically, SHAP insights demonstrated the critical impact of wind shears between 15 and 25 knots and the pronounced effects of complex terrain and sea breezes on significant turbulence occurrence, predominantly at altitudes below 1200 feet. These findings not only demonstrate the efficacy of the TabNet-SHAP model in enhancing aviation safety through improved turbulence prediction but also provide actionable insights for operational and safety protocols at airports prone to low-level wind shear.

Electric vehicles (EVs) have emerged as the pivotal strategy for mitigating carbon emissions. Nevertheless, the lack and imbalance of charging facilities have greatly reduced consumers’ willingness to buy EVs, thereby leading to stagnation in the EV market. Although many studies focus on the imbalance in the EV market, few of them have explored the pricing strategies of charging service providers from the perspective of consumers. Therefore, to assist charging service providers in optimizing pricing decisions, we propose a consumer-perspective utility model. Specially, this utility model analyzes both time and travel costs. Furthermore, considering three charging technologies including fixed charging, mobile charging vehicles, and mobile charging robots, we then apply game theory to characterize charging services providers competition scenarios between charging services providers. After theoretical and numerical analyses, we have the following five conclusions: 1) with the high time cost for consumers, the profits of fixed charging exceed those of mobile charging vehicles; 2) mobile charging robots intensify competition in the charging market and lower the price of conventional charging services; 3) mobile charging robots also reduce the demands for fixed charging services and expedite technological substitution; 4) cost of consumers time has a positive influence on mobile charging vehicles but a negative effect on other charging services; and 5) profits of the three charging providers increase with higher consumer time costs. Finally, we propose an innovative market analyzing tool based on research findings to assist charging service providers in their decision-making process.

To ensure the stable and safe operations, this paper presents a modeling framework of dynamic risk monitoring for chemical processes. Multi-source process data are firstly denoised by the Wavelet Transform (WT). The Spearman’s rank correlation coefficient (SRCC) of these data is calculated based on an appropriate time step and time window. An optimal correlation threshold is further applied to transform the SRCC matrix into an adjacency matrix. Accordingly, the model of complex networks (CNs) can be established for characterizing massive, disordered, and nonlinear process data. Network structure entropy is particularly introduced to transform process data into a single time series of relative risk. To illustrate its validity, a diesel hydrofining unit and Tennessee Eastman Process (TEP) are selected as test cases. Results show that the proposed modeling framework can effectively and reasonably monitor the risks of chemical processes in real time.

Pulmonary tuberculosis (PTB) is a major global health threat. Diagnosing PTB infectiousness is vital for clinical decision-making, but existing etiological examination methods do not meet the requirements for speed, accuracy, and cost effectiveness. Developing deep learning models for infectiousness identification based on computed tomography (CT) volume measurements holds promise for meeting these requirements. However, with limited samples and coarse annotations, the large amount of information in the CT volume poses a challenge for models to distinguish information related to patient-level labels, which often leads to severe model overfitting. In this study, A dual-branch framework is developed for identifying PTB infectiousness using CT volume. To address the issue of imbalance between the CT volume information and the patient-level labels, we propose a method for selecting representative slices to reduce redundant information and adopt a multiple-instance learning framework to improve label supervision. Furthermore, we incorporate multi-view projection information to compensate for the deficiency of global information caused by using single-dimensional slices as the input. Experimental results demonstrate that our strategy effectively mitigates overfitting and achieves desirable performance on an external test set, with an area under the receiver operating characteristic curve of 80.48%. This performance is superior to that obtained for models using the 3D CT volume or 2D projection images alone as the input.

The data-dependent hash methods are becoming more and more attractive because they perform well in fast retrieval and storing high-dimensional data. Most existing supervised hashes are centralized, such as supervised discrete hashing (SDH) and supervised discrete hashing with relaxation (SDHR). The SDH algorithm determines the regression target by using ordinary least squares regression and the traditional zero-one matrix encoding of class label information. And SDHR is a constraint to the regression target matrix so that each example is correctly classified and satisfies a larger margin, so as to achieve the purpose of optimizing the regression target. In real environment, a large amount of data will be distributed in different nodes. Therefore, the centralized hash method has great limitations. In this article, we propose distributed supervised discrete hashing algorithm with relaxation (DSDHR) based on SDHR. The SDHR algorithm is introduced into the distributed network. In this framework, all nodes share a centralized hash learning model. At the same time, in order to ensure that the distributed hash algorithm is updated in parallel on multiple nodes, consistency constraints are introduced. In each node, alternative iterative methods are used to obtain the binary hash code, regression target and hash function. Experiments show that DSDHR has certain competitive advantages over some centralized hashing algorithms and some distributed hashing algorithms.

The development of new media technology brings serious security problems to the transmission of secret remote sensing or military images. It is a new and challenging task to study the technology of protecting these secret images. In this paper, based on the powerful spatial feature extraction capability of the convolutional neural network, a novel two-channel deep hiding network (TDHN) is designed by introducing advanced ideas such as skip connection, feature fusion, etc., and the two channels are respectively used to input the cover image and the secret image simultaneously. This network consists of two parts: the hiding network and the extraction network. The sender uses the hiding network to hide a secret image in a common cover image and generates a hybrid image called the hidden image. The receiver uses the extraction network to extract and reconstruct the secret image from the hidden image. Meanwhile, an innovative loss function is constructed by introducing two metrics called MSE and SSIM. Experimental results show that the TDHN optimized by the loss function can generate the hidden image and extracted image in high quality. The SSIM value between the hidden image and the original cover image is up to around 0.99, and the SSIM value between the extracted image and the original secret image is up to around 0.98. Through testing on different datasets, it is verified that the designed and optimized TDHN has excellent generalization capability, and thus it has important theoretical significance and engineering value.

Elderly people generally suffer from diseases with cognitive decline, which seriously affect their daily lives. To assist these patients with cognitive decline, numerous cognitive aid devices have been researched and designed. Although these devices can achieve the aid tasks, they have to face two difficulties: low accuracy and long latency. In this paper, we present MobileAid, an aid system implemented on the mobile device, which assists cognitive decline patients to recognize objects. The key idea of this system is a two-step lightweight neural network design for the target recognition: context recognition and object recognition, and this design achieves low time delay. Considering the great success of convolutional neural networks in object recognition, we design a lightweight convolutional neural network by the combination of pooling layers and activation functions selection, which achieves high accuracy. Furthermore, we apply the depthwise separable convolution to reduce the resource consumption for deploying the system on mobile devices. The results of the extensive experiments we have conducted show that MobileAid can achieve 95% high accuracy and 90ms low time delay with few resource consumption.

This paper focuses on the problem of distributed adaptive estimation over dynamic multi-task networks, where a set of nodes is required to collectively estimate some parameters of interest from noisy measurements. Besides, since nodes in the network are constrained by communication power consumption and external interference in a non-stationary environment, the objective pursued by the node is prone to change or abnormality. The problem is worth considering in several contexts including multi-target tracking, multi-model classification and heterogeneous network segmentation. We propose a distributed adaptive clustering strategy, which is mainly composed of two procedures: normal task adaptation and the same task cluster. The task anomaly detection based on non-cooperative least-mean-squares (NC-LMS) algorithm and task switching detection based on diffusion maximum correntropy criterion (D-MCC) algorithm are provided. A series of scenarios, such as dynamic network, time-varying tasks and non-stationary (Gaussian and pulse interference) are simulated. We also discuss optimization schemes to design the NC-LMS and D-MCC weights and examine the estimate performance and clustering effects of the proposed algorithm by simulation results.

As the backbone of many real-world complex systems, networks interact with others in nontrivial ways from time to time. It is a challenging problem to detect subgraphs that have dependencies on each other across multiple networks. Instead of devising a method for a specific scenario, we propose a generic framework to discover subgraphs in multiple interdependent networks, which generalizes the classical subgraph detection problem in a single network and can be applied to more practical applications. Specifically, we propose the Graph Block-structured Gradient Hard Thresholding Pursuit (GB-GHTP) framework to optimize interdependent networks with block-structured constraints, which enjoys 1) a theoretical guarantee and 2) a nearly linear time complexity on the network size. It is demonstrated how our framework can be applied to three practical applications: 1) evolving anomalous subgraph detection in dynamic networks, 2) anomalous subgraph detection in networks of networks, and 3) connected dense subgraph detection in dual networks. We evaluate our framework on large-scale datasets with comprehensive experiments, which validate our framework's effectiveness and efficiency.

Electroencephalography (EEG) signal processing through imagery inputs using intelligent computation methods are practiced in the recent years for improving the accuracy of detecting neural disorders. Classification and analysis of the input imagery requires prior training and assisted error detection to improve the accuracy. In this article, pre-trained Gaussian mixture model (PGMM) is introduced for improving the accuracy of EEG signal imagery analysis. The proposed model relies on deep learning classifiers for analyzing the imagery using pixel based segmentation through pre-training models. The errors in classification are identified through recurrent convolution neural network training process as aided by the extracted features. Based on the pre-trained feature assessment, the false positive errors are mitigated to achieve a better accuracy (92%) under controlled classification time and high true positives.

In this paper, we propose a novel practical robust control algorithm for the Selective Compliance Articulated Robot Arm (SCARA) robot and verify the effectiveness through experiments. The dynamic model of the SCARA robot is established considering uncertainties, which include the nonlinear friction, parameter uncertainty, and external disturbance. To restrain the reversal chattering, we apply a modified Stribeck friction model with Gaussian compensation term as the friction description. The algorithm is composed of a proportional-derivative (PD) feedback term based on the model and a robust term. The formation of the robust part comprises the upper bound of the uncertainty. The Lyapunov minimax method is adopted to prove that the system is uniformly bounded and uniformly ultimately bounded, thus guaranteeing the practical stability of the system. Moreover, rapid controller prototyping cSPACE, as the experimental platform, can eliminate the tedious programming work and provide a great convenience for the experiments. The experimental results indicate that the robust control algorithm has good performance, which provides accurate trajectory tracking under the influence of uncertainties.

The hydraulic performance of centrifugal pumps is considerably affected by the impeller, and an effective optimization method for centrifugal pump impeller has been developed in the current study. A combination of the least-squares support vector regression machine (LSSVR) and particle swarm optimization algorithm (PSO) is proposed to redesign the impeller and improve the hydraulic performance. In the case study, four key geometric parameters of the impeller, namely, inlet angle, outlet angle, wrap angle and number of blades are selected for optimization. Maximum efficiency and constant head are selected as the optimization targets. During the optimization design, the required database for the LSSVR agent model is designed according to design of experiments. The optimal solution is then found in the established agent model space by the particle swarm algorithm and then verified by computational fluid dynamics. Ultimately, an improved impeller structure with an improved efficiency is provided. Numerical results show that the optimized impeller's efficiency is increased by 1.29% under the condition that the head is essentially unchanged. Then, the reason for the improvement of impeller hydraulic efficiency is explained by the entropy production method. The conclusions show that the PSO-LSSVR method can be used to optimize the pump impeller and achieve higher pump performance.

Excessive competition between taxis and subways has eroded the advantages of public transit systems such as worsening road traffic congestion and environment. This study aims to improve the appeal of subways by a comprehensive understating of competition between taxis and subways. We investigate competitive relationship between these two transportation modes by using empirical multi-source data. First, non-negative matrix factorization (NMF) algorithm is used to discover the spatiotemporal travel patterns of subway-competing taxi users (SCTUs). Second, we propose a new index to quantify the competitiveness of subways based on the actual mode choices results. Then, we reveal the spatiotemporal heterogeneity of competitiveness from perspective of subway network. Taking Beijing, China, for a case study, we extract a week's worth of GPS records on taxi trajectory and smartcard data of subways. Subway-competing taxi trips (SCTTs) account for the largest proportion of the total taxi trips. As a result, three basic patterns are found in SCTTs. Subway station pairs with high and less competition are divided according to competitiveness index. Among low competition station pairs, three spatial structures are observed, including low-competition collinearity corridors, radial communities, and links between paralleled subway lines. Combining the distribution results of travel pattern and competitiveness degree, short-term and long-term planning suggestions are recommended respectively for station pairs with high demand but low competitiveness and those with low demand and low competitiveness. These findings provide useful insights into promoting more effective and sensitive policies to balance the competition and attract more taxi passengers to the subway system.

Video transmission bandwidth can be substantially increased by concurrent transmission through multiple radio interfaces introduced by heterogeneous wireless networks. However, delivering high-quality real-time videos with stringent end-to-end delay constraints is still challenging due to the time-varying multipath channel status. Thus, to integrate available multipath bandwidth and guarantee low end-to-end delay, this paper investigates a real-time concurrent multipath video transmission scheme in heterogeneous access networks. First, we present a multipath buffer analysis model (MBAM) to derive the lower and upper bounds of the target video frame bits. Second, given the low-latency and received video quality smoothness, we propose a buffer-driven rate control and packet distribution (BRCPD) algorithm aided by frame-level rate control and multipath transmission buffer awareness based on MBAM. Finally, we implement the proposed application-layer solution BRCPD algorithm in a real-world testbed. The experiment results verify that different from existing multipath video transmission schemes, the proposed algorithm is able to leverage the MBAM-based frame-level rate control to guarantee the target low end-to-end delay. In particular, the BRCPD algorithm is able to guarantee an overdue loss rate of less than 5% under strictly limited delay constraints (250 ms) in the real-world testbed.

Cooperative vehicle infrastructure system (CVIS) is an essential part of the future intelligent transportation system (ITS). However, CVIS is uneasy to meet different QoS requirements of various information flows in the existing traffic environment. In fact, some real-time applications highly related to traffic safety early warning cannot be easily achieved or passengers' Internet accessing experience is always not very good. To overcome the above problem, this paper proposes a communication protocol named data networking for vehicle infrastructure cooperation (NDNVIC), which can respond to the vehicle's data query timely. In this paper, we utilize some ZigBee nodes as vehicle and RSU nodes and conduct several simulation experiments under a large-scale deployment with 150 RSU nodes to verify the effectiveness of NDNVIC. Compared with the traditional TCP/IP protocol, the response rate of NDNVIC is about four times, as fast as, that of TCP/IP.

Hash-based learning has attracted considerable attention due to its fast retrieval speed and low computational cost for the large-scale database. Compared with unsupervised hashing, supervised hashing achieves higher retrieval accuracy generally by leveraging supervised information. Most existing supervised hashing methods, such as supervised discrete hashing (SDH) and fast SDH (FSDH), are concerned more with the centralized setting. SDH regresses the hash code to its corresponding label, rather FSDH regressing each label to its corresponding hash code. However, in many realistic applications, large amounts of data are usually distributed across different sites. Thus, supervised distributed hashing (SupDisH), which is based on the distributed framework and supervised learning, has been proposed and liberates the limitations of centralized hashing. In this paper, based on FSDH, we propose the distributed fast supervised discrete hashing (DFSDH), which both inherits the excellent retrieval performance of SupDisH and gets significant enhancement in efficiency. Specifically, FSDH is introduced into a distributed framework, in which the centralized hash learning model is shared by all agents. Meanwhile, consistency constraints are introduced to ensure that multiple agents deal with distributed hash learning in parallel. For each agent, an alternate iterative procedure is employed to obtain high-quality binary codes and hashing function. The extensive experiments demonstrate that DFSDH is competitive to most centralized supervised hashing methods and existing distributed hashing methods.

With the continuous development of big data processing technology, data selection algorithms gradually attract the attention of researchers. In this paper, a distributed data-selection diffusion least mean square (DLMS) algorithm, which can improve the estimation accuracy of traditional data selection algorithms and can also censor data packets that do not bring enough innovation in wireless sensor networks, is proposed to censor the valid data in distributed iterative updates. And the adaption-then-combination strategy of the proposed algorithm is obtained. Meanwhile, in the distributed estimation system, the channel attacks are considered. When the network is under channel attacks, an adaptive credibility weight matrix is designed to improve the robustness of the distributed data-selection DLMS algorithm. We analyze the proposed algorithms in mean and mean-square performance. A series of simulations are carried out to demonstrate the effectiveness of the proposed algorithms. Moreover, it can be examined that through the comparison between the Metropolis weight strategy and the credibility weight strategy, the proposed credibility weight strategy is more robust in the face of channel attacks.

Short-term passenger flow forecasting is an essential component for the operation of urban rail transit (URT). Therefore, it is necessary to obtain a higher prediction precision with the development of URT. As artificial intelligence becomes increasingly prevalent, many prediction methods including the long short-term memory network (LSTM) in the deep learning field have been applied in road transportation systems, which can give critical insights for URT. First, we propose a novel two-step K-Means clustering model to capture not only the passenger flow variation trends but also the ridership volume characteristics. Then, a predictability assessment model is developed to recommend a reasonable time granularity interval to aggregate passenger flows. Based on the clustering results and the recommended time granularity interval, the LSTM model, which is called CB-LSTM model, is proposed to conduct short-term passenger flow forecasting. Results show that the prediction based on subway station clusters can not only avoid the complication of developing numerous models for each of the hundreds of stations, but also improve the prediction performance, which make it possible to predict short-term passenger flow on a network scale using limited dataset. The results provide critical insights for subway operators and transportation policymakers.

Handheld mobile photography is often affected by motion blur due to the difficulty of keeping the camera's stable. The existing processing method is usually a high-cost deblurring process of a computer, which seriously affects the user experience, and the deblurring effect is poor due to the lack of information on camera motion. Inspire by edging computing's ongoing efforts in automatically and collaboratively process more types of resources in the edge and cloud. In this paper, we present SID, a sensor-assisted image deblurring system for mobile devices. Using information about camera motion acquired from built-in sensors from smartphones (e.g. accelerometers, gyroscopes, and magnetometers), then estimating the point spread function, and combining the image's segmentation smoothing characteristics with spatial adaptation to image Deblurring. The image is preprocessed by the p-m nonlinear diffusion model, which preserves the characteristics of the image. The fuzzy image confirmation avoids the damage to the original high quality image. Wiener-Hoff optimization is used to optimize the effect of image restoration. We evaluated 400 photos with varying degrees of blur and size. Compared to traditional blind and unblind deconvolution methods, our algorithm shows significant advantages in both deblurring and processing delays.

Origin-destination (OD) forecasting is a difficult task in urban rail transit because many random factors can influence outcomes such as numerous OD pairs with few or even no flows. Therefore, treating all OD pairs equally, which is the method adopted by all existing studies, may not only increase model complexity and computation, but also negatively influence forecasting results. Therefore, in this study, we propose an indicator called OD attraction degree (ODAD) to address this problem in the field of OD forecasting. First, we introduce the ODAD indicator and five ODAD levels to describe the attraction between OD pairs. Based on the ODAD, an OD matrix pre-processing method is presented to prepare data for the LSTM model. Second, we use the mature long short-term memory (LSTM) network model to examine the effects of the introduction of ODAD. The LSTM model's advantage of dealing with variable-length sequences by leveraging the masking layer is creatively utilized. Finally, nine cases under different time granularities and different ODAD levels are thoroughly studied to explore their optimal combination. Based on this analysis, we recommend a time granularity of 30 min and an ODAD level of “Low” for actual subway operation. In this case, the root mean squared error, mean absolute error, and weighted mean absolute percentage error are 2.31%, 0.66%, and 27.28%, respectively, for a network of nearly 300 subway stations. The introduction of ODAD can provide critical insights for subway operators to conduct short-term OD forecasting.

Crowding force is an important index in evaluating the safety of those within high density crowds. However, studies on high density crowd simulation models have not been sufficiently developed, and related experimental studies are not sufficiently detailed to guide modeling research. Models that can reproduce real high density crowds and enable calibration of key parameters are lacking. In this study, we improved the typical social force model to simulate high density crowd motion with the calculation of crowding forces. To simulate high density crowds, pedestrians were modeled using the three-circle model, and an agent-based pedestrian model was formulated that can regulate the psychological force automatically according to crowd density. An individual crowd density calculation method is proposed that is more suitable for high density crowds. To calculate real crowding force, body stiffness was formulated and the corresponding parameters were calibrated by a series of experiments. Field experiments show that passengers bear greater force on their backs than the force on each shoulder in metro carriage. The modified model was verified by comparing the simulation results with observed data. We applied this model to the Beijing subway, and the theoretical maximum crowd densities in different seasons were calculated from the perspective of crowding force to ensure the safety of passengers, which are 7.2 people/m2 in winter and 10.3 people/m2 in summer. Therefore, to ensure passenger safety, the subway operation department should arrange more trains during winter peak hours.

Foresight is a manifestation of human-level intelligence. In model-based reinforcement learning, obtaining a perfect environment model is essential and significant. Although recent research has achieved impressive advances in improving prediction accuracy, existing environment models are unable to generalize well across environments, mainly because appearance-independent dynamics learning completely relies on appearance-specific representation learning. In this paper, we propose a novel predictive model named transferable environment model (TEM) which disentangles state representation and dynamics. The disentanglement allows the model to deal with different observation distributions and meanwhile share a cross-environment latent dynamics. In a 3D visual platform, we show that our model has good generalization performance in target environments with very few data. Furthermore, the TEM is able to continually adapt to a sequence of target environments without forgetting the knowledge for previous environments. To the best of our knowledge, this paper is the first to endow a predictive model with the ability to work across multiple environments.

Performance of a collaborative task is mostly dependent on the collective effort from participants. To accomplish a participatory task effectively and efficiently, the team formation problem (TFP) outweighs all other considerations. It is even more complicated when social connections among candidates is taken into account. As we can imagine, a large number of tasks require members of the team to be socially close. On the contrary, a portion of tasks, e.g., proposal review, pay more attention to a multidimensional view, and team members should be selected from a variety of cliques. Due to the nature of tasks, it is challenging to find a subset that meets the skill requirement of the task as well as socially diversity demand of team members from a pool of candidates. In this paper, we explore the TFP in a social network. Based on different task objectives, we first formulate the TFP as TFP with strong ties (TFP-ST) and TFP with weak ties (TFP-WT), respectively. Both TFP-ST and TFP-WT are proven to be NP-hard, and we then design corresponding heuristic algorithms to solve the two problems. Through extensive simulations, we show that the solution to TFP-ST can achieve significant improvement in terms of collaboration cost, team size, as well as running time, and the solution to TFP-WT can provide better performance than existing approaches at the same time.

Rising stars are junior individuals in the social network who will have high impacts with time accumulation. Rising star evaluation has become a research hotspot in network analysis area recently, which is helpful for decision support, resource allocation, and other practical problems. As a traditional social network, academic social network is stressed because of its heterogeneity and regular data structure. In this paper, we assume there are inside factors influencing individuals behaviors. We process the network parameters and mine inner factors via factor analysis, and train a decision tree to evaluate future impact. Experiment is processed on america physics society dataset, and the result shows our method has better performance than state-of-the-arts.

The popularity of campus information intelligence aims to provide better applications and services. However, we found that a phenomenon that has brought great inconvenience to students is queuing when pay in smart campus which needs to be resolved urgently. For example, buying meals in the school cafeteria at lunch or dinner time, paying school tuition, and so on. Also the privacy security is the primary prerequisite for smart campus's mobile electronic payment while transaction information can be intercepted easily by Hacker or hidden HD camera. In this paper, we propose the SafeMP, a mobile security and multi-to-one payment system employing the mechanical-wave transmission to reduce queue time and improve efficiency. The biggest conundrum with implementing this system is how to separate the multiple aliasing signals. The SafeMP is designed with a key finding: although the multiple aliasing signals are irregular, coupling degree between different mechanical wave signals is highly stable. We divide our work into three parts. First, we present a novel compensation method addressing the attenuation with the transmission distance. In addition, we define the degree of coupling to discriminate the aliasing for mechanical wave in ever slot. Finally, we design the rules of orthogonal coding to further connect the data bitstream of each time slot and restore the complete information which transmitted by different user, then the information separated can be corresponded to correct user. Our evaluation results show that the transmission distance does not influence the performance of the SafeMP. The SafeMP achieves an average accuracy of 97%.

Network coding is one of the key technologies to improve the throughput, efficiency, and reliability of two-way relay networks (TWRNs). However, it also brings some new security problems when the physical layer security is considered. In this paper, we present security network coding schemes with key generation from multipath channels to enhance the security performance for TWRN. First, a joint key generation (JKG) approach is proposed, which generates secret keys based on channel impulse responses of multipath channels. Different from the traditional approaches, key exchange is not necessary for the JKG approach such that the eavesdroppers cannot obtain any information about secret keys. Then, an adaptive quantization algorithm is proposed to adaptively choose different parameters for quantization in key generation, which can improve the secret key rate with an acceptable secret key disagreement probability and can still work even if the SNR is low. Finally, security network coding systems are constructed to combine the key generation approaches with the proposed algorithm. Simulation results show that the proposed schemes are valid and secure against wiretap attacks.

Rail wear inspection is vitally important in the railway industry. Conventional methods mainly use manual or static measurements to detect rail wear, which are inefficient, imprecise, and unreliable. To improve the accuracy and efficiency of rail wear inspection, a dynamic detection system based on a revised fast global registration (RFGR) algorithm was employed. First, the framework for online detection of rail wear with multi-profile was put forward to reduce the influence of vibrations of individual sections. Second, the RFGR method was proposed by using a robust weight function to convert the non-convex registration model to a convex problem, and the Levenberg-Marquardt method was used to solve nonlinear least-squares systems robustly. Finally, the Hausdorff distance was introduced to visualize the distance between the wear profile and the reference profile after alignment. The experimental results demonstrated that the RFGR algorithm was more accurate, robust, and effective than iterative closest point (ICP), sparse ICP, Vi's sparse ICP, and the fast global registration algorithm. For actual wear detection, the proposed method was more efficient and robust for the online dynamic detection of rail wear when compared with the single-profile-section-based inspection method.