The accurate prediction of traffic conditions, including speed, flow, and travel time, poses a critical challenge in urbanization that significantly impacts car owners and road administrators. However, in certain scenarios with restricted road data availability (e.g. lack of traffic light status and signal control strategies, cooperation between road administrators and third parties, etc.), it is imperative to make effective use of basic road information (e.g. historical traffic data and road connectivity) to improve both prediction accuracy and scalability on various arterial road networks against state-of-art deep learning models. In this paper, we propose a spatial-temporal learning framework TPGraph for an accurate prediction of arterial roads’ traffic data by effectively utilizing upstream and downstream road information. TPGraph is composed of three major parts: 1) A multi-scale temporal feature fusion module that utilizes a multi-head attention mechanism to integrate recently-periodic features, daily-periodic features, and weekly-periodic features; 2) A multi-graph convolution module that employs graph fusion and graph convolution networks to capture richer spatial semantics, and 3) A dynamic spatial-temporal prediction module that leverages a spatial-temporal transformer for single or multiple traffic-state predictions. Our proposed framework, TPGraph, leverages just multi-scale historical traffic conditions and readily accessible spatial factors as input to generate accurate predictions of future traffic conditions. We mainly evaluate the performance of our approach through multi-step prediction experiments conducted at hourly intervals, forecasting travel time or travel speed for each road at 15 mins, 30 mins, and 1 hour. Furthermore, we conduct extensive experiments on real-world arterial road datasets to demonstrate the superior predictive performance of TPGraph compared to existing methods.

Convolutional neural networks (CNNs) have achieved remarkable performance in driver drowsiness detection based on the extraction of deep features of drivers' faces. However, the performance of driver drowsiness detection methods decreases sharply when complications, such as illumination changes in the cab, occlusions and shadows on the driver's face, and variations in the driver's head pose, occur. In addition, current driver drowsiness detection methods are not capable of distinguishing between driver states, such as talking versus yawning or blinking versus closing eyes. Therefore, technical challenges remain in driver drowsiness detection. In this article, we propose a novel and robust two-stream spatial-temporal graph convolutional network (2s-STGCN) for driver drowsiness detection to solve the above-mentioned challenges. To take advantage of the spatial and temporal features of the input data, we use a facial landmark detection method to extract the driver's facial landmarks from real-time videos and then obtain the driver drowsiness detection result by 2s-STGCN. Unlike existing methods, our proposed method uses videos rather than consecutive video frames as processing units. This is the first effort to exploit these processing units in the field of driver drowsiness detection. Moreover, the two-stream framework not only models both the spatial and temporal features but also models both the first-order and second-order information simultaneously, thereby notably improving driver drowsiness detection. Extensive experiments have been performed on the yawn detection dataset (YawDD) and the National TsingHua University drowsy driver detection (NTHU-DDD) dataset. The experimental results validate the feasibility of the proposed method. This method achieves an average accuracy of 93.4% on the YawDD dataset and an average accuracy of 92.7% on the evaluation set of the NTHU-DDD dataset.

Hyperspectral images (HSIs) have gained high spectral resolution due to recent advances in spectral imaging technologies. This incurs problems, such as an increased data scale and an increased number of bands for HSIs, which results in a complex correlation between different bands. In the applications of remote sensing and earth observation, ground objects represented by each HSI pixel are composed of physical and chemical non-Euclidean structures, and HSI classification (HIC) is becoming a more challenging task. To solve the above problems, we propose a framework based on a deep attention graph convolutional network (DAGCN). Specifically, we first integrate an attention mechanism into the spectral similarity measurement to aggregate similar spectra. Therefore, we propose a new similarity measurement method, i.e., the mixed measurement of a kernel spectral angle mapper and spectral information divergence (KSAM-SID), to aggregate similar spectra. Considering the non-Euclidean structural characteristics of HSIs, we design deep graph convolutional networks (DeepGCNs) as a feature extraction method to extract deep abstract features and explore the internal relationship between HSI data. Finally, we dynamically update the attention graph adjacency matrix to adapt to the changes in each feature graph. Experiments on three standard HSI data sets, namely, the Indian Pines, Pavia University, and Salinas data sets, demonstrate that the DAGCN outperforms the baselines in terms of various evaluation criteria. For example, on the Indian Pines data set, the overall accuracy of the proposed method achieves 98.61% when the training sample is 10%.

In the field of hyperspectral image (HSI) classification, deep learning has helped achieve great successes. However, most of these achievements are made with very large amounts of labeled training data. Manual annotation of HSIs is labor intensive and time consuming. In practical HSI classification, there may only be a few labeled samples available. To perform HSI classification with a small number of labeled samples, a new few-shot classification model based on adaptive subspaces and featurewise transformation is proposed in this article. First, we design a 3-D local channel attention residual network to obtain the spatial–spectral features of HSIs. Then, a featurewise transformation strategy is introduced to enhance feature diversity to avoid model overfitting problems and to mitigate the impact of cross-domain problems. Finally, a subspace classifier is implemented to construct different subspace categories based on the embedded features of the limited labeled samples. Classification of an HSI sample is performed using spatial projection and a distance metric. The proposed model is trained using the metalearning mechanism to perform few-shot classification of HSIs. Four public datasets are utilized to construct a sufficient few-shot classification task named episodes for training. The other three public datasets are used to test the proposed model. Experiments show that our proposed method can outperform mainstream small sample HSI classification methods.

The graph-based hyperspectral image classification (HSIC) method has attracted wide attention because it can extract information with a non-Euclidean structure. Many graph-based HSIC works have achieved good results, but unresolved technical issues remain. For example, many graph nodes lead to high computational costs, and the mining of non-Euclidean structures is not sufficient. To solve these problems, we propose a graph attention network with an adaptive graph structure mining (GAT-AGSM) approach. Specifically, we first propose an HSIC framework with a superpixel feature subdivision (SFS) mechanism. In this framework, the number of nodes in the graph structure is reduced by using superpixel segmentation algorithms, and the SFS mechanism is designed to generate finer classification results. Second, we design the spatial–spectral attention layer with an adaptive graph structure mining (AGSM) mechanism for the graph attention network. The spatial–spectral attention layer can filter information in both spatial and spectral dimensions. The AGSM mechanism requires less manual intervention to dynamically generate non-Euclidean graph structures that better aggregate information. We conduct excessive experiments to compare the proposed GAT-AGSM with seven nongraph methods and three graph-based methods on widely used datasets. On the Indian Pines, Pavia University, and Salinas datasets, compared to the comparison method, the overall accuracy of GAT-AGSM is improved by at least 4.26%, 2.59%, and 1.41%, respectively. Experimental results show that GAT-AGSM has the best performance compared to the baselines in terms of various metrics.

Urban areas are negatively impacted by Carbon Dioxide (CO2 ) and Nitrogen Oxide (NOx) emissions. In order to achieve a cost-effective reduction of greenhouse gas emissions and to combat climate change, the European Union (EU) introduced an Emissions Trading System (ETS) where organizations can buy or receive emission allowances as needed. The current ETS is a centralized one, consisting of a set of complex rules. It is currently administered at the organizational level and is used for fixed-point sources of pollution such as factories, power plants, and refineries. However, the current ETS cannot efficiently cope with vehicle mobility, even though vehicles are one of the primary sources of CO2 and NOx emissions. In this study, we propose a new distributed Blockchain-based emissions allowance trading system called B-ETS. This system enables transparent and trustworthy data exchange as well as trading of allowances among vehicles, relying on vehicle-to-vehicle communication. In addition, we introduce an economic incentive-based mechanism that appeals to individual drivers and leads them to modify their driving behavior in order to reduce emissions. The efficiency of the proposed system is studied through extensive simulations, showing how increased vehicle connectivity can lead to a reduction of the emissions generated from those vehicles. We demonstrate that our method can be used for full life-cycle monitoring and fuel economy reporting. This leads us to conjecture that the proposed system could lead to important behavioral changes among the drivers

In many parts of the world, passengers traveling on underground metro systems do not enjoy uninterrupted Internet connectivity. This results in passenger frustration since during such trips the access of online social media services is a highly popular activity. Being the world's oldest underground metro system, London's underground is a typical transportation environment, where the Internet connectivity is often not available during journeys which predominantly take place underground along sub-surface and deep-level track lines. To alleviate the absence of continuous connectivity, we designed DeepOpp, a context-aware mobile system that facilitates offline access to online social media content. The DeepOpp operates efficiently due to its opportunistic approach: it executes content prefetching and caching operations when adequate urban 3G or WiFi signal is detected. The functionality of DeepOpp includes the crowdsourcing of measurements of signal characteristics (strength, bandwidth availability, and latency) which are subsequently used in predicting mobile network signal coverage and initiating data prefetching operations. During data prefetching, an optimization scheme selectively specifies the social media content to be cached based on current network conditions and device storage availability. We implemented DeepOpp as an Android application which we trialled during real trips on the London underground. Our evaluations show that the DeepOpp offers significant reduction when compared with existing approaches in terms of power usage and the volume of data downloaded. Even though we only tested DeepOpp in the London underground metro system, its feature set makes it readily applicable in similar underground metro systems (in cities like New York, Paris, and Shanghai) as well as in situations, where mobile device users suffer from significant connectivity interruptions.

The scarcity of parking spaces in cities leads to a high demand for timely information about their availability. In this paper, we propose a crowdsensed parking system, namely ParkCrowd, to aggregate on-street and roadside parking space information reliably, and to disseminate this information to drivers in a timely manner. Our system not only collects and disseminates basic information, such as parking hours and price, but also provides drivers with information on the real time and future availability of parking spaces based on aggregated crowd knowledge. To improve the reliability of the information being disseminated, we dynamically evaluate the knowledge of crowd workers based on the veracity of their answers to a series of location-dependent point of interest control questions. We propose a logistic regression-based method to evaluate the reliability of crowd knowledge for real-time parking space information. In addition, a joint probabilistic estimator is employed to infer the future availability of parking spaces based on crowdsensed knowledge. Moreover, to incentivise wider participation of crowd workers, a reliability-based incentivisation method is proposed to reward workers according to their reliability and expertise levels. The efficacy of ParkCrowd for aggregation and the dissemination of parking space information has been evaluated in both real-world tests and simulations. Our results show that the ParkCrowd system is able to accurately identify the reliability level of the crowdsensed information, estimate the potential availability of parking spaces with high accuracy, and be successful in encouraging the participation of more reliable crowd workers by offering them higher monetary rewards.

Accessing online social media content on underground metro systems is a challenge due to the fact that passengers often lose connectivity for large parts of their commute. As the oldest metro system in the world, the London underground represents a typical transportation network with intermittent Internet connectivity. To deal with disruption in connectivity along the sub-surface and deep-level underground lines on the London underground, we have designed a context-aware mobile system called DeepOpp that enables efficient offline access to online social media by prefetching and caching content opportunistically when signal availability is detected. DeepOpp can measure, crowdsource and predict signal characteristics such as strength, bandwidth and latency; it can use these predictions of mobile network signal to activate prefetching, and then employ an optimization routine to determine which social content should be cached in the system given real-time network conditions and device capacities. DeepOpp has been implemented as an Android application and tested on the London Underground; it shows significant improvement over existing approaches, e.g. reducing the amount of power needed to prefetch social media items by 2.5 times. While we use the London Underground to test our system, it is equally applicable in New York, Paris, Madrid, Shanghai, or any other urban underground metro system, or indeed in any situation in which users experience long breaks in connectivity.

We present ADDSEN middleware as a holistic solution for Adaptive Data processing and dissemination for Drone swarms in urban SENsing. To efficiently process sensed data in the middleware, we have proposed a cyber-physical sensing framework using partially ordered knowledge sharing for distributed knowledge management in drone swarms. A reinforcement learning dissemination strategy is implemented in the framework. ADDSEN uses online learning techniques to adaptively balance the broadcast rate and knowledge loss rate periodically. The learned broadcast rate is adapted by executing state transitions during the process of online learning. A strategy function guides state transitions, incorporating a set of variables to reflect changes in link status. In addition, we design a cooperative dissemination method for the task of balancing storage and energy allocation in drone swarms. We implemented ADDSEN in our cyber-physical sensing framework, and evaluation results show that it can achieve both maximal adaptive data processing and dissemination performance, presenting better results than other commonly used dissemination protocols such as periodic, uniform and neighbor protocols in both single-swarm and multi-swarm cases.

Open WiFi access points (APs) are demonstrating that they can provide opportunistic data services to moving vehicles. We present CrowdWiFi, a novel system to look up roadside WiFi APs located outdoors or inside buildings. CrowdWiFi consists of two components: online compressive sensing (CS) and offline crowdsourcing. Online CS presents an efficient framework for the coarse-grained estimation of nearby APs along the driving route, where received signal strength (RSS) values are recorded at runtime, and the number and location of the APs are recovered immediately based on limited RSS readings and adaptive CS operations. Offline crowdsourcing assigns the online CS tasks to crowd-vehicles and aggregates answers on a bipartite graphical model. Crowd-server also iteratively infers the reliability of each crowd-vehicle from the aggregated sensing results, and then refines the estimation of the APs using weighted centroid processing. Extensive simulation results and real testbed experiments confirm that CrowdWiFi can successfully reduce the computation cost and energy consumption of roadside WiFi lookup, while maintaining satisfactory localization accuracy.

Roadside units (RSUs) are public and personal wireless access points that can provide communications with infrastructure in ad hoc vehicular networks. We present CLOCS (Counting and Localization using Online Compressive Sensing), a novel system to retrieve both the number and locations of RSUs through war-driving. CLOCS employs online compressive sensing (CS), where received signal strength (RSS) values are recorded at runtime, and the number and location of RSUs are recovered immediately based on limited RSS readings. CLOCS also uses fine retrieval based on an expectation maximization method along the driving route. Extensive simulation results and experiments in a real testbed deployed in the campus of the University of California, Irvine, confirm that CLOCS can successfully reduce the number of measurements for RSU recovery, while maintaining satisfactory counting and localization accuracy. In addition, data dissemination, time cost, and effects of different mobile scenarios using CLOCS are analyzed, and the impact of CLOCS on network connectivity is studied using Microsoft VanLan traces.

We propose a voted dual averaging method for on- line classification problems with explicit regularization. This method employs the update rule of the regularized dual averaging (RDA) method proposed by Xiao, but only on the subsequence of training examples where a classification error is made. We derive a bound on the number of mistakes made by this method on the training set, as well as its generalization error rate. We also intro- duce the concept of relative strength of regularization, and show how it affects the mistake bound and gener- alization performance. We examine the method using l1-regularization on a large-scale natural language pro- cessing task, and obtained state-of-the-art classification performance with fairly sparse models.

In this paper, we present CrowdWiFi, a novel vehicular middleware to identify and localize roadside WiFi APs that are located outside or inside buildings. Our work is motivated by the recent surge in availability of open WiFi access points (APs) that are enabling opportunistic data services to moving vehicles. Two key elements of CrowdWiFi that provide vehicles with opportunistic WiFi access include (a) an online compressive sensing component and (b) an offline crowdsourcing module. Online compressive sensing (CS) techniques are primarily used to for the coarse-grained estimation of nearby APs along the driving route; here, the received signal strength (RSS) values are recorded at runtime, and the number and locations of APs are recovered immediately based on limited RSS readings. The offline crowdsourcing mechanism assigns the online CS tasks to crowd-vehicles and aggregates answers using a bipartite graphical model. This offline crowdsourcing executes at a crowd-server that iteratively infers the reliability of each crowd-vehicle from the aggregated sensing results and refines the estimation of APs using weighted centroid processing. Extensive simulation results and real testbed experiments confirm that CrowdWiFi can successfully reduce the number of measurements needed for AP recovery, while maintaining satisfactory counting and localization accuracy. In addition, the impact of CrowdWiFi middleware on WiFi handoff and data transmission applications is examined.

It is a challenge to design efficient routing protocols for vehicular ad hoc networks (VANETs) because of their highly dynamic properties. We address the vehicular communication problem in urban hybrid networks and present a hybrid routing scheme for data dissemination in VANETs. Location-based crowdsourcing of nearby roadside units (RSUs) has been applied to the infrastructural support of inter-vehicle, vehicle-to-roadside, and inter-roadside communications in hybrid VANETs. The combination of RSU resources and ad hoc networks involves an online probabilistic RSU retrieval algorithm that uses coarse- and fine-grained localization to estimate the number and location of available RSUs; a network coding based multicast routing for dense VANETs using maximum distance separation (MDS) code and local topology information from the forwarding set to achieve robust communication and max-flow min-cut data dissemination; an application of opportunistic routing, using a carry-and-forward scheme to solve the forwarding disconnection problem in sparse VANETs; and a routing switch mechanism to guarantee quality of service (QoS) under various network connectivity and deployment configurations. The performance of our hybrid routing scheme is evaluated using both simulations and real testbed experiments.

A primary goal of intelligent transportation systems (ITS) is to improve road safety. The ability for vehicles to communicate is a promising way to alleviate traffic accidents by reducing the response time associated with human reaction to nearby drivers. In addition the limitations of standard driving can be overcome by providing drivers with instantaneous information about complications up ahead. Shockwaves, induced by vehicle speed differentials, are a typical mobility pattern that occurs with the formation and propagation of vehicle queues. These induce sudden braking and increase the occurrence of traffic incidents. In this paper, we investigate safety applications in highways with shockwave mobility and different lane configurations in vehicular ad hoc networks (VANET). We evaluate the performance of multi-hop broadcast communication using the ns-2 simulator with vehicles following a shockwave mobility pattern in fully-connected traffic streams. We propose mechanism to improve broadcast reliability using dynamic transmission range that leverages our understanding of fundamental traffic flow relationships.

Inter-vehicle communication is a promising way to share and disseminate real-time and nearby safety information on the road. However, several pressing open questions require solutions in order to achieve high reliability and efficiency with these systems. Further, previous studies show that mobility model can significantly influence the communication performance in vehicular networks. In this paper, we analyze communication in stop-and-go waves and propose a method to optimize an important network parameter, the transmission range, based on traffic pattern measures. Our findings suggest a transmission range adjustment scheme that achieves high reliability by considering network coverage and packet reception rates.