Search

This paper proposes a cooperative strategy of connected and automated vehicles (CAVs) longitudinal control for a mixed connected and automated traffic environment based on deep reinforcement learning (DRL) algorithm, which enhances the string stability of mixed traffic, car following efficiency, and energy efficiency. Since the sequences of mixed traffic are combinatory, to reduce the training dimension and alleviate communication burdens, we decomposed mixed traffic into multiple subsystems where each subsystem is comprised of human-driven vehicles (HDV) followed by cooperative CAVs. Based on that, a cooperative CAV control strategy is developed based on a deep reinforcement learning algorithm, enabling CAVs to learn the leading HDV’s characteristics and make longitudinal control decisions cooperatively to improve the performance of each subsystem locally and consequently enhance performance for the whole mixed traffic flow. For training, a distributed proximal policy optimization is applied to ensure the training convergence of the proposed DRL. To verify the effectiveness of the proposed method, simulated experiments are conducted, which shows the performance of our proposed model has a great generalization capability of dampening oscillations, fulfilling the car following and energy-saving tasks efficiently under different penetration rates and various leading HDVs behaviors.

Connected Automated Vehicles (CAVs) bring promise of increasing traffic capacity and energy efficiency by forming platoons with short headways on the road. However at low CAV penetration, the capacity gain will be small because the CAVs that randomly enter the road will be sparsely distributed, diminishing the probability of forming long platoons. Many researchers propose to solve this issue by platoon organization strategies, where the CAVs search for other CAVs on the road and change lanes if necessary to form longer platoons. However, the current literature does not analyze a potential risk of platoon organization in disrupting the flow and reducing the capacity by inducing more lane changes. In this research, we use driving model of Cooperative Adaptive Cruise Control (CACC) vehicles and human-driven vehicles that are validated with field experiments and find that platoon organization can indeed drop the capacity with more lane changes. But when the traffic demand is well below capacity, platoon organization forms longer CAV platoons without reducing the flow. Based on this finding, we develop the Flow-Aware platoon organization strategy, where the CAVs perform platoon organization conditionally on the local traffic state, i.e., a low flow and a high speed. We simulate the Flow-Aware platoon organization on a realistic freeway network and show that the CAVs successfully form longer platoons, while ensuring a maximal traffic flow.

Traffic flow measurement is very important for traffic management systems. However, the existing traditional measurement approaches are highly time-consuming and expensive to continuously gather the required data and to maintain the corresponding equipment, such as loop detectors and video cameras. On the other hand, many services on the web propose to estimate automobile travel time taking into account traffic conditions thanks to crowd sourced data (Floating Car Data). This work proposes to reconstruct, from estimated travel time, traffic flows using machine learning method. In particular, we evaluate the capacity of Gaussian Process Regressor (GPR) to address this issue. After obtaining estimated travel time on a given route, a clustering process shows that travel duration profiles in each day can be associated to different “types of day”. Then, different regressors are trained in order to estimate traffic flows from travel duration. In the “multi-model” variant, we trained a Regressor for each type of day. Conversely, in the “single model” variant, only one Regressor is trained (the type of day is not taken into account). This is an innovative work to estimate and reconstruct the traffic flow in transportation networks with machine learning method from aggregated Floating Car Data (FCD). A series of experiments are conducted to compare the estimated traffic flows, obtained by the proposed single model and multi-model, and the real ones from actual sensors. The obtained results show that both single model and multi-models can capture the tendency of real traffic flows. Furthermore, the performance can be improved by regulating parameters in GPR machine learning model, such as half width of sample window and sample size (a whole week or only weekdays), and multi-models can highly increase the performance compared with the single model. Therefore, the proposed GPR machine learning and FCD based new method can replace those traditional loop detectors for the measurement of traffic flow.

Geographic information systems, global positioning systems, traffic flow sensors and cellular phones are sources of real-time traffic data in road networks. However, many vehicle routing algorithms do not account for this real-time information. In this paper, we consider the problem of adjusting in real-time a time-dependent delivery plan to respond to dynamic changes in travel times. We also consider a variant of the problem in which some customer requests can be canceled. The goal is to minimize disruption by maintaining as much as possible the current planned routes, although without compromising too much solution quality. Computational results obtained by solving instances with up to 500 customers are reported and compared with a strategy that maintains the planned routes, whatever the cost.

This paper presents a methodology for developing demand-responsive passenger flow control strategies for oversaturated metro networks. A stated preference-off-revealed preference survey was conducted to capture passengers’ behavioural responses to the passenger flow control strategies, and revealed significant behavioural change intentions in terms of departure times and mode choices. Currently, such behavioural responses are often ignored, leading to deviations in estimating the size of a target group and performance of a proposed strategy. To address this issue, in this study, a mathematical programming with equilibrium constraints (MPEC) approach was developed to optimise demand-responsive passenger flow control strategies. The proposed MPEC model highlights the importance of balancing the operational efficiency of a metro system with service fairness perceived by passengers. In particular, a nested logit-based stochastic user equilibrium problem was included to accommodate potential changes in demand patterns driven by candidate passenger flow control strategies. Two empirical cases based on the Guangzhou and Beijing metros were used to demonstrate the effectiveness of the proposed model and solution algorithm. The results show that the inclusion of service fairness considerations does not contradict the pursuit of efficiency. Instead, dual emphases on service fairness and passengers’ behavioural responses help create a win–win situation for both metro operators and passengers. In the Guangzhou and Beijing metros, the highest section load rate decreased from 126.6% to 105.7% and from 122.98% to 106.87%, respectively, under the optimal passenger flow control strategies. The corresponding network load Gini coefficients improved from 0.278 to 0.259 and from 0.269 to 0.248, demonstrating the remarkable performance of this approach in regards to peak cutting and load balancing.

Dynamic vehicular origin–destination (OD) flow is a fundamental component of traffic network modeling and its estimation has long been studied. Although ideal observing conditions and behavioral assumptions are often indispensable for estimation, day-to-day traffic recurrences and variations are seldom utilized to improve the estimation performance. In this paper, we propose a new method to recover day-to-day dynamic OD flows using both connected vehicle (CV) trajectories and automatic vehicle identification (AVI) observations. The method involves two modules: the first module provides reliable prior OD flows given limited observations, while the second module seeks the optimal estimates based on the prior OD flows. In the first module, linear projection is extended to consider temporal and spatial variation of the CV penetration rate, and non-negative Tucker decomposition (NTD) is adopted to address the data sparsity issue caused by the low CV penetration rate. In the second module, a self-supervised learning model called the latency-constrained autoencoder (LCAE) is established to search for the optimal OD flows according to the priors with given robust latent features. To avoid local minima and ensure consistency between estimates, a novel algorithm called adaptive sub-sample correction (ASC) is proposed and integrated into the optimization process of LCAE, which can iteratively correct the most inconsistent samples based on the day-to-day traffic flow characteristics. The proposed method is examined on an empirical urban arterial network, a calibrated simulation network, and a synthetic large-scale grid network. Our results indicated that the proposed method requires very few AVI detectors and CV trajectories to achieve competitive estimation performance against two benchmark models. Furthermore, general robustness to several factors with respect to observing conditions and data quality was investigated, and satisfactory scalability was also demonstrated in terms of both estimation accuracy and computational cost.

Traffic management on (smart) motorways is becoming increasingly important to extract the most out of the limited capacity that is available, especially during peak periods. Accurate modelling of traffic management measures in traffic flow simulation is therefore paramount to make accurate short, and/or long-term predictions, to aid offline and online applications. In this work we propose a novel macroscopic network loading model that supports variable speed limits (VSL). Most existing aggregate network loading models that support VSL impose speed limits instantaneously across the entire road section, ignoring the fact that only vehicles passing the speed limit sign are aware of the new limit (either by a driver seeing the signal, or in case of autonomous/connected vehicles via infrastructure-to vehicle-technology), this leads to a discrepancy between model and reality that one ideally would want to avoid. Another challenge that arises is that current state-of-the-art macroscopic network loading models, such as the link transmission model (LTM), do not track the internal state of the link while constructing their solution. This is especially useful in large-scale (macroscopic) simulations where its increased efficiency is valued highly. Unfortunately, this desirable property makes modelling traffic management measures such as VSL much more challenging. Virtually all existing methods to date rely on tracking the link’s internal state to support VSL. Therefore, very few methods exist that offer proper support for VSL in LTM. At the same time, the methods that do exist have various limitations. In this work we propose a continuous-time LTM model that address these limitations, yields exact solutions, and supports VSL. The proposed solution method is an extension of the event-based LTM of Bliemer and Raadsen (2019) and Raadsen and Bliemer (2019) which can be considered a special case with a single speed limit. Three case studies are provided to demonstrate feasibility and general applicability of the proposed method and solution algorithm.

This paper aims to answer the research question as to optimal design of decision-making processes for autonomous vehicles (AVs), including dynamical selection of driving velocity and route choices on a transportation network. Dynamic traffic assignment (DTA) has been widely used to model travelers's route choice or/and departure-time choice and predict dynamic traffic flow evolution in the short term. However, the existing DTA models do not explicitly describe one's selection of driving velocity on a road link. Driving velocity choice may not be crucial for modeling the movement of human drivers but it is a must-have control to maneuver AVs. In this paper, we aim to develop a game-theoretic model to solve for AVs's optimal driving strategies of velocity control in the interior of a road link and route choice at a junction node. To this end, we will first reinterpret the DTA problem as an N-car differential game and show that this game can be tackled with a general mean field game-theoretic framework. The developed mean field game is challenging to solve because of the forward and backward structure for velocity control and the complementarity conditions for route choice. An efficient algorithm is developed to address these challenges. The model and the algorithm are illustrated on the Braess network and the OW network with a single destination. On the Braess network, we first compare the LWR based DTA model with the proposed game and find that the driving and routing control navigates AVs with overall lower costs. We then compare the total travel cost without and with the middle link and find that the Braess paradox may still arise under certain conditions. We also test our proposed model and solution algorithm on the OW network., 32 pages, 13 figures

This study proposes a multi-vehicle anticipation-based discrete-continuous choice modelling framework for describing driver behaviour in heterogeneous disordered traffic (HDT) conditions. To incorporate multi-vehicle anticipation, the concept of an influence zone around a vehicle (subject vehicle) is introduced. Vehicles within the influence zone can potentially influence the subject vehicle’s driving behaviour. Further, driving decisions are characterized as combination of discrete and continuous components. The discrete component involves the decision to accelerate, decelerate, or maintain constant speed and the continuous component involves the decision of how much to accelerate or decelerate. A copula-based joint modelling framework that allows dependencies between discrete and continuous components is proposed. Such a joint modelling framework recognizes that the discrete and continuous decisions are made simultaneously, and common unobserved factors influence both decisions. Additionally, truncated distributions are employed for the continuous model components to avoid the prediction of unrealistically high acceleration or deceleration values. The parameters of the proposed model are estimated using a trajectory dataset from Chennai, India. The empirical results underscore (a) the importance of considering multi-vehicle anticipation for describing driving behaviour in HDT conditions, and (b) the efficacy of the joint discrete-continuous system for modelling driving behaviour. Further, not all traffic environment variables found to influence the discrete decisions were found influential on continuous decisions and vice versa. Moreover, the influence of several variables was found to be stronger on the decision to accelerate or decelerate than on the decision of how much to accelerate or decelerate.

Improving the system performance of a traffic network by dynamically controlling the routes of connected and automated vehicles (CAVs) is an appealing profit that CAVs can bring to our society. Considering that there may be a long way to achieve 100% CAV penetration, we discuss in this paper the mixed traffic flow of human driven vehicles (HDVs) and CAVs on a transportation network. We first propose a double queue (DQ) based mixed traffic flow model to describe the link dynamics as well as the flow transitions at junctions. Based on this mixed flow model, we develop a dynamic bi-level framework to capture the behavior and interaction of HDVs and CAVs. This results in an optimal control problem with equilibrium constraints (OCPEC), where HDVs’ route choice behavior is modeled at the lower level by the instantaneous dynamic user equilibrium (IDUE) principle and the CAVs’ route choice is modelled by the dynamic system optimal (DSO) principle at the upper level. We show how to discretize the OCPEC to a mathematical programming with equilibrium constraints (MPEC) and discuss its properties and solution techniques. The non-convex and non-smooth properties of the MPEC make it hard to be efficiently solved. To overcome this disadvantage, we develop a decomposition based heuristic model predictive control (HMPC) method by decomposing the original MPEC problem into two separate problems: one IDUE problem for HDVs and one DSO problem for CAVs. The experiment results show that, compared with the scenario that all vehicles are HDVs, the proposed methods can significantly improve the network performance under the mixed traffic flow of HDVs and CAVs.

This study develops a Complete LAne-Changing Decision (CLACD) modeling framework that explains both the mandatory and discretionary lane-changing behaviors, incorporates both the successful and failed lane-changing attempts, and describes lane-changing behaviors in both the traditional and connected environments. The connected environment provides real-time and valuable information about surrounding vehicles and downstream situations (e.g., congestion ahead) that drivers cannot foresee. Such information can be particularly useful for complex driving maneuvers like discretionary lane-changing (DLC), which requires information about surrounding traffic speeds and subsequent gaps available on the adjacent lanes. DLC modeling efforts in the connected environment appear to be nascent, primarily because of: (a) the novelty of the connected environment (and the consequent scarcity of the relevant data), and (b) the lack of a behaviorally sound DLC model even for the traditional environment (that is, without driver-aid messages). Moreover, the existing models omit failed DLC attempts by motorists that are likely to create traffic disruptions and have a significant impact on traffic safety. To overcome these challenges, this study develops a DLC model (i.e., CLACD model) for the traditional environment, which is extended for the connected environment. An integrated approach is employed to model DLC behavior by combining the target lane selection using the utility theory approach and the gap acceptance behavior using a game theory approach. To collect high-quality vehicle trajectory data in the connected environment, the CARRS-Q advanced driving simulator was used. The CLACD models (for both the traditional and connected environments) are calibrated using NGSIM and the driving simulator data, and the CLACD models’ performances were rigorously assessed using various performance indicators. Results reveal that the CLACD models can effectively capture the observed DLC decisions with high accuracy. The performance of CLACD models is compared with the performance of two alternative models described in the literature, revealing that the CLACD models outperform the alternatives. A sensitivity analysis suggests a consistent trend of the CLACD models, which agrees with past research. Furthermore, the CLACD model shows a realistic prediction of traffic flow patterns compared with the utility theory model. By incorporating failed DLC attempts into the CLACD models, the performance and realism of CLACD models have been significantly improved.

The study proposes a mathematical framework to explain the stochastic behavioural patterns of personal mobility (PM) devices under low-carbon heterogeneous traffic conditions in shared lanes. We create a set of anticipation factors in a stochastic PM behaviour model to tackle sensitivities to both space headway and relative speed against intra- and inter-modes. The proposed behaviour model involves a deterministic and a stochastic force. In the deterministic force, the anticipation factors are used in an optimal velocity model and a full velocity difference model. In the stochastic force, the Langevin equation is used to capture PMs’ stochastic characteristics against movements of other PMs, pedestrians, and bicycles, and the effect of lateral interactions. We carried out real-world circular experiments of mixed sustainable modes to verify the performance of the proposed models. Five models’ performances are compared under four different traffic conditions, including bike-mixed, pedestrian-mixed, low-speed, and high-speed conditions. We confirmed that newly created anticipation factors play a significant role in all models under all conditions to partially influence the following PM devices’ behaviour from the leading two different sustainable modes. The validation results illustrate the excellence of the proposed method. Consequently, behavioural uncertainty is well captured by the stochastic PM devices following models under all traffic conditions, although it requires more parameters than the deterministic PM behavioural models. The proposed method paves the way for the stochastic CF model’s applicability to describe PM devices’ behavioural dynamics under mixed traffic conditions using anticipation factors. Besides, it lays the foundation stone of PM devices’ dynamics in a shared lane to construct effective regulations and safety standards.