Zhuang Dai

Bio: Zhuang Dai is an academic researcher from Beihang University. The author has contributed to research in topics: Autoencoder & Deep learning. The author has an hindex of 5, co-authored 6 publications receiving 1186 citations.

Learning Traffic as Images: A Deep Convolutional Neural Network for Large-Scale Transportation Network Speed Prediction.

Abstract: This paper proposes a convolutional neural network (CNN)-based method that learns traffic as images and predicts large-scale, network-wide traffic speed with a high accuracy. Spatiotemporal traffic dynamics are converted to images describing the time and space relations of traffic flow via a two-dimensional time-space matrix. A CNN is applied to the image following two consecutive steps: abstract traffic feature extraction and network-wide traffic speed prediction. The effectiveness of the proposed method is evaluated by taking two real-world transportation networks, the second ring road and north-east transportation network in Beijing, as examples, and comparing the method with four prevailing algorithms, namely, ordinary least squares, k-nearest neighbors, artificial neural network, and random forest, and three deep learning architectures, namely, stacked autoencoder, recurrent neural network, and long-short-term memory network. The results show that the proposed method outperforms other algorithms by an average accuracy improvement of 42.91% within an acceptable execution time. The CNN can train the model in a reasonable time and, thus, is suitable for large-scale transportation networks.

Learning Traffic as Images: A Deep Convolutional Neural Network for Large-Scale Transportation Network Speed Prediction

Abstract: This paper proposes a convolutional neural network (CNN)-based method that learns traffic as images and predicts large-scale, network-wide traffic speed with a high accuracy. Spatiotemporal traffic dynamics are converted to images describing the time and space relations of traffic flow via a two-dimensional time-space matrix. A CNN is applied to the image following two consecutive steps: abstract traffic feature extraction and network-wide traffic speed prediction. The effectiveness of the proposed method is evaluated by taking two real-world transportation networks, the second ring road and north-east transportation network in Beijing, as examples, and comparing the method with four prevailing algorithms, namely, ordinary least squares, k-nearest neighbors, artificial neural network, and random forest, and three deep learning architectures, namely, stacked autoencoder, recurrent neural network, and long-short-term memory network. The results show that the proposed method outperforms other algorithms by an average accuracy improvement of 42.91% within an acceptable execution time. The CNN can train the model in a reasonable time and, thus, is suitable for large-scale transportation networks.

Joint optimization of scheduling and capacity for mixed traffic with autonomous and human-driven buses: A dynamic programming approach

Abstract: It is a common practice for transit lines with fluctuating passenger demands to use demand-driven bus scheduling to reduce passenger waiting time and avoid bus overcrowding. However, current literature on the demand-driven bus scheduling generally assumes fixed bus capacity and exclusively optimizes bus dispatch headways. With the advent of connected and autonomous vehicle technology and the introduction of autonomous minibus/shuttle, the joint design of bus capacity and dispatch headway holds promises to further improving the system efficiency while reducing operating and passenger costs. This paper formulates this problem as an integer nonlinear programming model for transit systems operating with mixed human-driven and autonomous buses. In such mixed operating environment, the model simultaneously considers: (1) dynamic capacity design of autonomous bus, i.e., autonomous buses with varying capacity can be obtained by assembling and/or dissembling multiple autonomous minibuses; (2) trajectory control of autonomous bus, i.e., autonomous bus can dynamically adjust its running time as a function of its forward and backward headways; and (3) stop-level passenger boarding and alighting behavior. The objective of the model is designed to balance the trade-off between the operating costs of dispatching different types of bus and the costs of increased passenger waiting time due to inadequate bus dispatching. The model is solved using a dynamic programming approach. We show that the proposed model is effective in reducing passenger waiting time and total operating cost. Sensitivity analysis is further conducted to explore the impact of miscellaneous factors on optimal dispatching decisions, such as penetration rate of autonomous bus, bus running time variation, and passenger demand level.

A predictive headway-based bus-holding strategy with dynamic control point selection: A cooperative game theory approach

Abstract: Bus holding is a widely used control method to regularize bus headways and reduce bus bunching. The method works in such a way by delaying buses at control points if their departure times or headways deviate from the planned ones. However, it may result in reduced bus commercial speeds and increased passenger onboard travel time. To avoid this problem, researchers have suggested that control points be spaced cautiously along the route such that only a few are needed. This study proposes a predictive headway-based bus holding strategy with dynamic control point importance ranking and selection based on the cooperative game theory. The framework considers not only individual control points’ impact but also the collective group control effects. Specifically, the proposed framework consists of two components: a performance model and a cooperative game model. The performance model predicts headway performances of all running buses when different control point combinations are in effect. Dynamic bus running times and passenger demands are reflected in the model. Then, these headway performances are passed to the cooperative game model with control points being players and improvements in headway performances compared with that under no holding control being the utility function. The game is solved by Myerson value, a concept that extends Shapley value used for the normal cooperative game and considers the cooperation structure and potential worth of coalitions. We use Myerson value to rank the importance of control points on regularizing headways, as it measures the average marginal utility contribution of a control point to all possible coalitions that exclude that point. We prove that Myerson value lies in the Ω-core of the game and thus satisfies allocation efficiency, individual and coalition rationality. The proposed framework is applied to target headway control and two-way-looking self-equalizing headway control. Simulation results show that the framework can significantly reduce passenger waiting time and bus headway variation.

Diffusion Convolutional Recurrent Neural Network: Data-Driven Traffic Forecasting

Abstract: Spatiotemporal forecasting has various applications in neuroscience, climate and transportation domain. Traffic forecasting is one canonical example of such learning task. The task is challenging due to (1) complex spatial dependency on road networks, (2) non-linear temporal dynamics with changing road conditions and (3) inherent difficulty of long-term forecasting. To address these challenges, we propose to model the traffic flow as a diffusion process on a directed graph and introduce Diffusion Convolutional Recurrent Neural Network (DCRNN), a deep learning framework for traffic forecasting that incorporates both spatial and temporal dependency in the traffic flow. Specifically, DCRNN captures the spatial dependency using bidirectional random walks on the graph, and the temporal dependency using the encoder-decoder architecture with scheduled sampling. We evaluate the framework on two real-world large scale road network traffic datasets and observe consistent improvement of 12% - 15% over state-of-the-art baselines.

Traffic Graph Convolutional Recurrent Neural Network: A Deep Learning Framework for Network-Scale Traffic Learning and Forecasting

Abstract: Traffic forecasting is a particularly challenging application of spatiotemporal forecasting, due to the time-varying traffic patterns and the complicated spatial dependencies on road networks. To address this challenge, we learn the traffic network as a graph and propose a novel deep learning framework, Traffic Graph Convolutional Long Short-Term Memory Neural Network (TGC-LSTM), to learn the interactions between roadways in the traffic network and forecast the network-wide traffic state. We define the traffic graph convolution based on the physical network topology. The relationship between the proposed traffic graph convolution and the spectral graph convolution is also discussed. An L1-norm on graph convolution weights and an L2-norm on graph convolution features are added to the model’s loss function to enhance the interpretability of the proposed model. Experimental results show that the proposed model outperforms baseline methods on two real-world traffic state datasets. The visualization of the graph convolution weights indicates that the proposed framework can recognize the most influential road segments in real-world traffic networks.

Spatiotemporal Multi-Graph Convolution Network for Ride-Hailing Demand Forecasting

Abstract: Region-level demand forecasting is an essential task in ridehailing services. Accurate ride-hailing demand forecasting can guide vehicle dispatching, improve vehicle utilization, reduce the wait-time, and mitigate traffic congestion. This task is challenging due to the complicated spatiotemporal dependencies among regions. Existing approaches mainly focus on modeling the Euclidean correlations among spatially adjacent regions while we observe that non-Euclidean pair-wise correlations among possibly distant regions are also critical for accurate forecasting. In this paper, we propose the spatiotemporal multi-graph convolution network (ST-MGCN), a novel deep learning model for ride-hailing demand forecasting. We first encode the non-Euclidean pair-wise correlations among regions into multiple graphs and then explicitly model these correlations using multi-graph convolution. To utilize the global contextual information in modeling the temporal correlation, we further propose contextual gated recurrent neural network which augments recurrent neural network with a contextual-aware gating mechanism to re-weights different historical observations. We evaluate the proposed model on two real-world large scale ride-hailing demand datasets and observe consistent improvement of more than 10% over stateof-the-art baselines.

Short-term forecasting of passenger demand under on-demand ride services: A spatio-temporal deep learning approach

Abstract: Short-term passenger demand forecasting is of great importance to the on-demand ride service platform, which can incentivize vacant cars moving from over-supply regions to over-demand regions. The spatial dependencies, temporal dependencies, and exogenous dependencies need to be considered simultaneously, however, which makes short-term passenger demand forecasting challenging. We propose a novel deep learning (DL) approach, named the fusion convolutional long short-term memory network (FCL-Net), to address these three dependencies within one end-to-end learning architecture. The model is stacked and fused by multiple convolutional long short-term memory (LSTM) layers, standard LSTM layers, and convolutional layers. The fusion of convolutional techniques and the LSTM network enables the proposed DL approach to better capture the spatio-temporal characteristics and correlations of explanatory variables. A tailored spatially aggregated random forest is employed to rank the importance of the explanatory variables. The ranking is then used for feature selection. The proposed DL approach is applied to the short-term forecasting of passenger demand under an on-demand ride service platform in Hangzhou, China. The experimental results, validated on the real-world data provided by DiDi Chuxing, show that the FCL-Net achieves the better predictive performance than traditional approaches including both classical time-series prediction models and state-of-art machine learning algorithms (e.g., artificial neural network, XGBoost, LSTM and CNN). Furthermore, the consideration of exogenous variables in addition to the passenger demand itself, such as the travel time rate, time-of-day, day-of-week, and weather conditions, is proven to be promising, since they reduce the root mean squared error (RMSE) by 48.3%. It is also interesting to find that the feature selection reduces 24.4% in the training time and leads to only the 1.8% loss in the forecasting accuracy measured by RMSE in the proposed model. This paper is one of the first DL studies to forecast the short-term passenger demand of an on-demand ride service platform by examining the spatio-temporal correlations.