Freeway ramp merging control in the mixed traffic consisting of connected and automated vehicles (CAVs) and human-driven vehicles (HDVs) is one of the bottlenecks in the development of autonomous driving (AD) technologies due to complex multi-vehicle interactions. To address this, we propose a decentralized human-like control framework to help CAVs merge smoothly and interact more effectively with HDVs in a mixed-traffic environment. First, a stochastic potential game model is proposed to characterize the uncertainty of HDVs’ actions and optimize individual actions while considering their impact on the global system. A model predictive controller (MPC) is designed to minimize the accumulated cost of the potential function over a time horizon. Next, a distributed algorithm is developed to solve the proposed optimization problem in parallel while reducing the computational burden. To capture the characteristics of human driving behavior and help CAVs take more human-like actions, we calibrate the parameters in the proposed model using a real-world trajectory dataset. The performance of the proposed method is tested in a realistic two-lane merging zone scenario. Experimental results show that the proposed method enables CAVs to merge smoothly and ultimately improves traffic efficiency in merging zones. Additionally, the solution quickly converges to the optimal result using the proposed distributed algorithm, supporting its application in decentralized AD systems.

Traffic congestion in urban areas presents a major challenge to efficient transportation systems. Recent advancements in traffic management provide promising solutions, with perimeter control emerging as a technique to tackle network-wide congestion. However, it is crucial to identify geographically connected homogeneously congested areas for effective implementation. This research explores the application of clustering techniques, particularly geographical self-organizing maps (GeoSOM), to identify spatially connected and homogeneously congested areas within transportation networks. While GeoSOM has found applications across various domains, its adaptation to transportation networks for congestion clustering is novel. This study introduces and implements an adaptation of the GeoSOM algorithm tailored for the large-scale urban environment of downtown Los Angeles. Its performance is assessed through a comparative evaluation with two other clustering algorithms, namely DBSCAN and K-means. The results demonstrate that GeoSOM surpasses other clustering algorithms, exhibiting improvements of up to 43% in traffic density variance, up to 61% in the spatial quantization error, and 15% in the quantization error. This finding demonstrates that the proposed clustering algorithm is effective in identifying a spatially homogeneous congested area within a large-scale transportation network.

Recently, efforts have been made to standardize signal phase and timing (SPaT) messages. These messages contain signal phase timings of all signalized intersection approaches. This information can thus be used for efficient motion planning, resulting in more homogeneous traffic flows and uniform speed profiles. Despite efforts to provide robust predictions for semi-actuated signal control systems, predicting signal phase timings for fully-actuated controls remains challenging. This paper proposes a time series prediction framework using aggregated traffic signal and loop detector data. We utilize state-of-the-art machine learning models to predict future signal phases’ duration. The performance of a Linear Regression (LR), Random Forest (RF), a light gradient-boosting machine (LightGBM), a bidirectional Long-Short-Term-Memory neural network (BiLSTM) and a Temporal Convolutional Network (TCOV) are assessed against a naive baseline model. Results based on an empirical data set from a fully-actuated signal control system in Zurich, Switzerland, show that state of the art machine learning models outperform conventional prediction methods.

Abstract A better understanding of Driving Patterns and their relationship with geographical driving areas could bring great benefits for smart cities, including the identification of good driving practices for saving fuel and reducing carbon emissions and accidents. The process of extracting driving patterns can be challenging due to issues such as the collection of valid data, clustering of population groups, and definition of similar behaviors. Naturalistic Driving methods provide a solution by allowing the collection of exhaustive datasets in quantitative and qualitative terms. However, exploiting and analyzing these datasets is complex and resource-intensive. Moreover, most of the previous studies, have constrained the great potential of naturalistic driving datasets to very specific situations, events, and/or road sections. In this paper, we propose a novel methodology for extracting driving patterns from naturalistic driving data, even from small population samples. We use Geographic Information Systems (GIS), so we can evaluate drivers’ behavior and reactions to certain events or road sections, and compare across situations using different spatial scales. To that end, we analyze some kinematic parameters such as speeds, acceleration, braking, and other forces that define a driving attitude. Our method favors an adequate mapping of complete datasets enabling us to achieve a comprehensive perspective of driving performance.

Since the introduction of the Macroscopic Fundamental Diagram (MFD), many traffic control strategies and algorithms have been developed to implement MFD-based perimeter control over a specific urban region. A model-based controller consists of two components: aplant modelthat represents reality; and aprediction modelused to determine optimal control actions. In most studies, the authors assume a constant average trip length for all drivers traveling within the same region, for theprediction model. In these studies about perimeter control and MFD traffic models, the controllers show a good performance because accumulations, i.e. traffic states, from the plant are used to reflect the initial state of the prediction model with a high frequency (about a few seconds). However, this average trip length changes over time as it depends on the Origin-Destination flow decomposition, playing an important role in real applications. The main contributions of this paper are twofold. First, we show that the assumption about constant trip lengths used in theprediction modeldeteriorates the controller’s performance for low frequency updates of the optimal control actions. Second, we propose a methodological framework based on the Unscented Kalman Filter (UKF) for dynamically adjusting the average trip lengths and accumulations. Our test results on a real city network show that applying this methodological framework significantly improves the controller’s performance.

Understanding the inter-relationships between traffic flow, density, and speed through the study of the fundamental diagram of road traffic is critical for traffic modelling and management. Consequently, over the last 85 years, a wealth of models have been developed for its functional form. However, there has been no clear answer as to which model is the most appropriate for observed (i.e. empirical) fundamental diagrams and under which conditions. A lack of data has been partly to blame. Motivated by shortcomings in previous reviews, we first present a comprehensive literature review on modelling the functional form of empirical fundamental diagrams. We then perform fits of 50 previously proposed models to a high quality sample of 10 150 empirical fundamental diagrams pertaining to 25 cities. Comparing the fits using information criteria, we find that the non-parametric Sun model greatly outperforms all of the other models. The Sun model maintains its winning position regardless of road type and congestion level. Our study, the first of its kind when considering the number of models tested and the amount of data used, finally provides a definitive answer to the question “Which model for the functional form of an empirical fundamental diagram iscurrentlythe best?”. The word “currently” in this question is key, because previously proposed models adopt an inappropriate Gaussian noise model with constant variance. We advocate that future research should shift focus to exploring more sophisticated noise models. This will lead to an improved understanding of empirical fundamental diagrams and their underlying functional forms.

We propose a kinematic wave-based Deep Convolutional Neural Network (Deep CNN) to estimate high-resolution traffic speed fields from sparse probe vehicle trajectories. We introduce two key approaches that allow us to incorporate kinematic wave theory principles to improve the robustness of existing learning-based estimation methods. First, we propose an anisotropic traffic kernel for the Deep CNN. The anisotropic kernel explicitly accounts for space-time correlations in macroscopic traffic and effectively reduces the number of trainable parameters in the Deep CNN model. Second, we propose to use simulated data for training the Deep CNN. Using a targeted simulated data for training provides an implicit way to impose desirable traffic physical features on the learning model. In the experiments, we highlight the benefits of using anisotropic kernels and evaluate the transferability of the trained model to real-world traffic using the Next Generation Simulation (NGSIM) and the German Highway Drone (HighD) datasets. The results demonstrate that anisotropic kernels significantly reduce model complexity and model over-fitting, and improve the physical correctness of the estimated speed fields. We find that model complexity scales linearly with problem size for anisotropic kernels compared to quadratic scaling for isotropic kernels. Furthermore, evaluation on real-world datasets shows acceptable performance, which establishes that simulation-based training is a viable surrogate to learning from real-world data. Finally, a comparison with standard estimation techniques shows the superior estimation accuracy of the proposed method.

This paper proposes a convex optimization based algorithm for queue profile estimation in a connected vehicle environment, which can also be used for trajectory reconstruction, delay evaluation, etc. This algorithm generalizes the widely-adopted assumption of a linear back of queue (BoQ) curve to a piecewise linear BoQ curve to consider more practical scenarios. The piecewise linear BoQ curve is estimated via a convex optimization model, ensuring efficient computation. Moreover, this paper explicitly handles cases with low penetration rates and low sampling rates, as well as measurement noises. In addition, the proposed methodology is extended to an urban arterial, reusing the estimated departure information from the upstream intersections to further improve the estimation accuracy. Finally, two online implementation approaches are presented to perform real-time queue estimation. The proposed methodology is tested with two datasets: the Lankershim data set in the NGSIM project and the simulated dataset of Wehntalerstrasse, Zürich, Switzerland. Results show that the error is less than 1.5 cars in undersaturated scenarios and 5.2 cars in oversaturated scenarios if the penetration rates are larger than 0.1 and sampling rates are higher than 0.05 s-1. It is demonstrated that by considering a piecewise linear BoQ curve, the estimation accuracy can be improved by up to 16%. Incorporating flow successfully can also reduce the estimation error by up to 16%. Results further show that the proposed methodology is robust to measurement errors. It is finally shown that the proposed framework can be solved within a reasonable time (0.8 s), which is sufficient for most real-time applications.

The platooning of trucks can be considered to be a potential approach to mitigate some of the negative effects that trucking can have on traffic streams. This paper proposes a cooperative distributed approach for forming/modifying platoons of trucks based on consensus algorithms. In this approach, trucks exchange information about their current status in real time, and the platoon is formed in consecutive iterations. This distributed consensus-based algorithm is compared with a centralized optimization-based algorithm for truck platooning, in which the trucks move with a set of predetermined speeds for a definite amount of time to form a platoon. The two approaches are tested and compared using various scenarios generated based on real data collected on a highway in Basel, Switzerland. Based on the results, the consensus-based algorithm proved to be a more general scheme that is able to form platoons even in cases with large initial separation of trucks. This algorithm is able to handle complex situations using its capability to form partial platoons.

One of the main challenges arising when calibrating a complex traffic simulation model concerns the selection of the most important input parameters. The quasi-optimized trajectory-based elementary effects (quasi-OTEE) and the Kriging-based sensitivity analysis (SA) are two recently developed efficient approaches for the SA of computationally expensive simulation models. In this paper, two experimental studies using two different traffic simulation models (i.e., Aimsun and VISSIM) are presented to compare these two approaches and to better understand their advantages and disadvantages. Results show that both approaches are able to identify, to a good degree, the important parameters. In particular, the quasi-OTEE is better for screening the parameters, whereas the Kriging-based SA has higher precision in ranking the parameters. These findings suggest the following rule of thumb for the SA of computationally expensive traffic simulation models: the quasi-OTEE SA can be used first to screen the parameters and to decide which parameters to discard. Then, the Kriging-based SA can be used to refine the analysis and calculate first-order indexes to identify the correct rank of the important parameters.