Disturbance, which is generally unknown to the controller, is unavoidable in real-world systems and it may affect the expected system state and output. Existing control methods, like robust model ...predictive control, can produce robust solutions to maintain the system stability. However, these robust methods trade the solution optimality for stability. In this article, a method called generative adversarial control networks (GACNs) is proposed to train a controller via demonstrations of the optimal controller. By formulating the optimal control problem in the presence of disturbance, the controller trained by GACNs obtains neuro-optimal solutions without knowing the future disturbance and determines the objective function explicitly. A joint loss, composed of the adversarial loss and the least square loss, is designed to be used in the training of the generator. Experimental results on simulated systems with disturbance show that GACNs outperform other compared control methods.
Air pollution presents a serious health challenge in urban metropolises. While accurately monitoring and forecasting air pollution are highly crucial, existing data-driven models have yet fully ...captured the complex interactions between the temporal characteristics of air pollution and the spatial characteristics of urban dynamics. Our proposed Deep-AIR fills this gap to provide fine-grained city-wide air pollution estimation and station-wide forecast, by exploiting domain-specific features (including Air Pollution, Weather, Urban Morphology, Transport, and Time-sensitive features), with a hybrid CNN-LSTM structure to capture the spatio-temporal features, and <inline-formula> <tex-math notation="LaTeX">1\times 1 </tex-math></inline-formula> convolution layers to enhance the learning of temporal and spatial interaction. Deep-AIR outperforms compatible baselines by a higher accuracy of 1.5%, 2.7%, and 2.3% for Hong Kong and 1.4%, 1.4% and 3.3% for Beijing in fine-grained 1-hr pollution estimation, and 1-hr and 24-hr forecasts, respectively. Saliency analysis reveals that for Hong Kong, spatial features, including street canyon and road density, are the best predictors for NO 2 , while temporal features, including historical air pollutants and weather, are the best predictors for PM 2.5 . For Beijing, historical air pollutant data, traffic congestion, wind direction and seasonal indicator are the best predictors for all pollutants. PM 10 in Hong Kong is achieving the best estimation and forecast accuracy, whilst CO in Beijing is achieving the best results.
Income inequality presents a key challenge to urban sustainability across the developed economies. Traditionally, accurate high granularity income data are generally obtained from field surveys. ...However, due to privacy considerations, field subjects are hesitant to provide accurate personal income data. A
(SSIG) model is thereby developed to estimate district-based high granularity income for New York City (NYC). As compared to the state-of-the-art Gaussian Processes (GP) income estimation model based entirely on spatial information, SSIG incorporates socio-economic domain-specific knowledge into a GP model. For SSIG to be explainable, SHapley Additive exPlanations (SHAP) analysis is undertaken to evaluate the relative contribution of various key individual socio-economic variables to district-based per-capita and median household income in NYC. Differentiating from traditional income inequality studies based predominantly on linear or log-linear regression model, SSIG presents a novel income-based model architecture, capable of modelling complex non-linear relationships. In parallel, SHAP analysis serves an effective analytical tool for identifying the key attributes to income inequality. Results have shown that SSIG surpasses other state-of-the-art baselines in estimation accuracy, as far as per-capita and median household income estimation at the Tract-level and the ZIP-level in NYC are concerned. SHAP results have indicated that having a bachelor or a postgraduate degree can accurately predict income in NYC, despite that between-district income inequality due to Sex/Race remains prevalent. SHAP has further confirmed that between-district income gap is more associated with Race than Sex. Furthermore, ablation study shows that socio-economic information is more predictive of income at the ZIP-level, relative to the spatial information. This study carries significant implications for policy-making in a developed context. To promote urban economic sustainability in NYC, policymakers can attend to the growing income disparity (income inequality) contributed by Sex and Race, while giving more higher education opportunities to residents in the lower-income districts, as the estimated per-capita income is more sensitive to the proportion of adults ≥25 holding a bachelor's degree. Finally, interpretative SHAP analysis is useful for investigating the relative contribution of socio-economic inputs to any predicted outputs in future machine-learning-driven socio-economic analyses.
This paper presents a novel delay aware synchrophasor recovery and prediction framework to address the problem of missing power system state variables due to the existence of communication latency. ...This capability is particularly essential for dynamic power system scenarios where fast remedial control actions are required due to system events or faults. While a wide area measurement system can sample high-frequency system states with phasor measurement units, the control center cannot obtain them in real-time due to latency and data loss. In this work, a synchrophasor recovery and prediction framework and its practical implementation are proposed to recover the current system state and predict future states utilizing existing incomplete synchrophasor data. The framework establishes an iterative prediction scheme, and the proposed implementation adopts recent machine learning advances in data processing. Simulation results indicate the superior accuracy and speed of the proposed framework, and investigations are made to study its sensitivity to various communication delay patterns for pragmatic applications.
Vehicular ad hoc network (VANET) is an emerging type of networks which facilitates vehicles on roads to communicate for driving safety. The basic idea is to allow arbitrary vehicles to broadcast ad ...hoc messages (e.g. traffic accidents) to other vehicles. However, this raises the concern of security and privacy. Messages should be signed and verified before they are trusted while the real identity of vehicles should not be revealed, but traceable by authorized party. Existing solutions either rely heavily on a tamper-proof hardware device, or cannot satisfy the privacy requirement and do not have an effective message verification scheme. In this paper, we provide a software-based solution which makes use of only two shared secrets to satisfy the privacy requirement (with security analysis) and gives lower message overhead and at least 45% higher successful rate than previous solutions in the message verification phase using the bloom filter and the binary search techniques (through simulation study). We also provide the first group communication protocol to allow vehicles to authenticate and securely communicate with others in a group of known vehicles.
A CNN-LSTM Model for Traffic Speed Prediction Cao, Miaomiao; Li, Victor O. K.; Chan, Vincent W. S.
2020 IEEE 91st Vehicular Technology Conference (VTC2020-Spring)
Conference Proceeding
Increasingly serious traffic congestion requires an accurate and timely traffic speed prediction, which will significantly benefit both individual drivers and decision makers in travel planning and ...traffic management. However, traffic speed prediction is a long-standing and challenging topic. Due to the availability of traffic datasets and powerful computation resources, deep learning becomes a promising solution to this problem. In this paper, based on Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) models, we propose a model named CLM, which is the first to make use of CNN to extract the features of daily and weekly periodicity of traffic speed at the target area and also extract the spatiotemporal features together with the output of CNN by LSTM layers. We conduct comprehensive simulations to assess the performance of our proposed method based on the real-world dataset of Hong Kong. The results indicate that our proposed CLM model can better predict traffic speed in different forecast time periods than the other five competing methods, including SVR, MLP, Lasso, Random forest, and LSTM.
•A data-driven model is developed to estimate air quality at the fine-grained CA level.•PM2.5 is positively correlated with SD in HK with statistical significance.•Human-, location- and justice-based ...policies can better address PM2.5 induced SD/EI.
Poor air quality has extremely detrimental health consequences, including cancer, stroke, asthma or heart disease. Existing research on air pollution-induced environmental injustice (EI) in Hong Kong (HK) is based on sparse air pollution data due to the limited number of pollution monitoring stations, rendering the study of the relationship between air pollution exposure and social deprivation (SD), and the subsequent study of EI at finer geographical scales difficult. A key research question remains to be explored in a methodologically rigorous manner: Are the socially deprived exposed to a higher level of PM2.5 pollution in HK? Our study develops a Granger Causality model, utilizing ubiquitous urban dynamics closely related with air pollution, including ambient pollutants concentration, traffic, meteorology and urban morphology to provide a fine-grained estimation of air pollution in HK at 100m×100m spatial resolution, achieving a 82% accuracy. We focus specifically on ambient PM2.5 concentration, given its serious health consequences and the preliminary evidence of differential PM2.5 distribution across different socio-economic groups in HK. We investigate if there is any differential distribution in PM2.5 pollution across people residing in constituency areas (CAs) with different levels of SD. In our study, SD is measured by the Social Deprivation Index (SDI), which is a composite indicator comprising four socio-economic status variables, namely, low-income, low-education, non-professional occupation, and non-owner occupier, selected and combined via principal component analysis. We conclude there is a statistically significant, positive relationship between ambient PM2.5 concentration and SDI in HK, based on the SDI and mean PM2.5 exposure values derived from 412 CAs (R2=1.4%, p-value <0.01, based on ordinary least squares), justifying the existence of PM2.5-induced SD and EI at the CA level in HK. Our study highlights an emerging need for HK to develop more integrated, human-centric, location- and justice-based environmental policies, and the need to adopt evidence-based policy-decision-making to properly address air pollution-induced EI. Our policy implications and recommendations can be extended to the rest of the world, particularly the Asian metropolis, as well as places where GDP growth is rapid, population density and pollution concentration (including particulate pollution) appear to be high, and the income gap between the rich and the poor is widening.
Attempts have been made to estimate PM 2.5 and PM 10 values from smartphone images, given that deploying highly accurate air pollution monitors throughout a city is a highly expensive undertaking. ...Departing from previous machine learning studies which primarily focus on pollutant estimation based on single day-time images, our proposed deep learning model integrates Residual Network (ResNet) with Long Short-Term Memory (LSTM), extracting spatial-temporal features of sequential images taken from smartphones instead for estimating PM 2.5 and PM 10 values of a particular location at a particular time. Our methodology is as follows: First, we calibrated two small portable air quality sensors using the reference instruments placed in the official air quality monitoring station, located at Central, Hong Kong (HK). Second, we verified experimentally that any PM 2.5 and PM 10 values obtained via our calibrated sensors remain constant within a radius of 500 meters. Third, 3024 outdoor day-time and night-time images of the same building were taken and labelled with corresponding PM 2.5 and PM 10 ground truth values obtained via the calibrated sensors. Fourth, the proposed ResNet-LSTM was constructed and extended by incorporating meteorological information and one short path. Results have shown that, as compared to the best baselines, ResNet-LSTM has achieved 6.56% and 6.74% reduction in MAE and SMAPE for PM 2.5 estimation, and 13.25% and 11.03% reduction in MAE and SMAPE for PM 10 estimation, respectively. Further, after incorporating domain-specific meteorological features and one short path, Met-ResNet-LSTM-SP has achieved the best performance, with 24.25% and 20.17% reduction in MAE and SMAPE for PM 2.5 estimation, and 28.06% and 24.57% reduction in MAE and SMAPE for PM 10 estimation, respectively. In future, our deep-learning image-based air pollution estimation study will incorporate sequential images obtained from 24-hr operating traffic surveillance cameras distributed across all parts of the city in HK, to provide full-day and more fine-grained image-based air pollution estimation for the city.
Facial emotions are expressed through a combination of facial muscle movements, namely, the Facial Action Units (FAUs). FAU intensity estimation aims to estimate the intensity of a set of ...structurally dependent FAUs. Contrary to the existing works that focus on improving FAU intensity estimation performance, this study investigates how knowledge distillation (KD) incorporated into a training model can improve FAU intensity estimation efficiency while achieving the comparable level of performance. Given the intrinsic structural characteristics of FAU, it is desirable to distill deep structural relationships, namely, DSR-FAU, using heatmap regression. Our methodology is as follows: First, a feature map-level distillation loss is applied to ensure that the student network and the teacher network share similar feature distributions. Second, the region-wise and channel-wise relationship distillation loss functions are introduced to penalize the difference in structural relationships. Specifically, the region-wise relationship can be represented by the structural correlations across the facial features, whereas the channel-wise relationship is represented by the implicit FAU co-occurrence dependencies. Third, we compare the model performance of DSR-FAU with the state-of-the-art models, based on two benchmarking datasets. It is shown that our model achieves comparable performance, with a lower number of model parameters and lower computation complexities.
PDF
