Accurate short-term traffic flow forecasting facilitates active traffic control and trip planning. Most existing traffic flow models fail to make full use of the temporal and spatial features of traffic data. This study proposes a short-term traffic flow prediction model based on a convolution neural network (CNN) deep learning framework. In the proposed framework, the optimal input data time lags and amounts of spatial data are determined by a spatio-temporal feature selection algorithm (STFSA), and selected spatio-temporal traffic flow features are extracted from actual data and converted into a two-dimensional matrix. The CNN then learns these features to construct a predictive model. The effectiveness of the proposed method is evaluated by comparing the forecast results with actual traffic data. Other existing models are also evaluated for comparison. The proposed method outperforms baseline models in terms of accuracy.

With the rapid development of sensor technologies, time series data collected by multiple and spatially distributed sensors have been widely used in different research fields. Examples of such data include geo-tagged temperature data collected by temperature sensors, air pollutant monitoring data, and traffic data collected by road traffic sensors. Due to sensor failure, communication errors and storage loss, etc., data collected by sensors inevitably includes missing data. However, models commonly used in the analysis of such large-scale data often rely on complete data sets. This paper proposes a model for the imputation of missing data of traffic flow, which combines a self-attention mechanism, an auto-encoder, and a generative adversarial network, into a self-attention generative adversarial imputation net (SA-GAIN). The introduction of the self-attention mechanism can help the proposed model to effectively capture correlations between spatially-distributed sensors at different time points. Adversarial training through two neural networks, called generators and discriminators, allows the proposed model to generate imputed data close to the real data. In comparison with different imputation models, the proposed model shows the best performance in imputing missing data.

Traffic congestion at signalized intersections often leads to serious impacts on adjacent intersections on a corridor. To enhance intersections’ throughput efficiency, traffic signals are commonly coordinated across intersections. Traditional signal coordination methods control the adjacent intersections by setting a fixed phase offset. However, these traditional coordination methods may have poor adaptability to dynamic traffic conditions, which can cause additional congestion. To reduce arterial traffic delays, this paper develops an adaptive coordination control method based on multi-agent reinforcement learning (MARL). Most existing MARL-based methods rely on impractical assumptions to improve their performance in complex and dynamic traffic scenarios. To overcome these assumptions, this paper proposes a fully scalable MARL algorithm for arterial traffic signal coordination based on the proximal policy optimization algorithm. We apply a parameter-sharing training protocol to mitigate the slow convergence due to nonstationarity and to reduce computational requirements. In addition, a new action setting is designed by using the lead-lag phase sequence to simultaneously improve the implementation and coordination flexibility of the method. Extensive simulation experiments and comparisons with existing methods demonstrate that the proposed method performed stably in both simulated and real-world arterial corridors. Hence, the proposed signal coordination method can alleviate traffic congestion more effectively than existing traditional and MARL-based methods.

Ship collision is the most common type of accident in the Northern Baltic Sea, posing a risk to the safety of maritime transportation. Near miss detection from automatic identification system (AIS) data provides insight into maritime transportation safety. Collision risk always triggers a ship to maneuver for safe passing. Some frenetic rudder actions occur at the last moment before ship collision. However, the relationship between ship behavior and collision risk is not fully clarified. Therefore, this work proposes a novel method to improve near miss detection by analyzing ship behavior characteristic during the encounter process. The impact from the ship attributes (including ship size, type, and maneuverability), perceived risk of a navigator, traffic complexity, and traffic rule are considered to obtain insights into the ship behavior. The risk severity of the detected near miss is further quantified into four levels. This proposed method is then applied to traffic data from the Northern Baltic Sea. The promising results of near miss detection and the model validity test suggest that this work contributes to the development of preventive measures in maritime management to enhance to navigational safety, such as setting a precautionary area in the hotspot areas. Several advantages and limitations of the presented method for near miss detection are discussed.

The rapid development of traffic theory and information technology has provided diversified and large-scale traffic data resources for traffic research and urban traffic management. At the same time, these data also present many challenges, such as missing data and deviations in data collection. Many researchers have reported that inaccurate or incomplete measurements of traffic variables can be corrected based on either traditional traffic flow theory, which ignores the randomness of traffic, or are performed using machine learning methods, which emphasize data quantity, but do not make effective use of domain knowledge. This paper proposes a Traffic Factor State Network framework defined by traffic factors and their links to represent the relationships between traffic factors; this framework includes not only obvious traffic factors like volume and speed, but also hidden traffic factors such as the environmental impact factor, which is a variable used to represent complex road conditions. This variable is used to describe the influence of non-traffic flow parameters such as road condition and environmental factors, and is estimated by the EM (Expectation Maximization) algorithm based on historical data. This study used a high-order multivariate Markov model to implement the TFSN, which was then used to establish a stochastic model of speed and related factors. A large amount of historical data was used to calculate and calibrate the strength of the links between the model factors. Finally, a stochastic model of speed prediction was established. The verification results compared with actual cases demonstrate the validity and applicability of the proposed model.