Transportation Research Part C-Emerging Technologies最新文献

Perceived risk is crucial in designing trustworthy and acceptable vehicle automation systems. However, our understanding of perceived risk dynamics remains limited, and corresponding computational models are scarce. This study formulates a new computational perceived risk model based on potential collision avoidance difficulty (PCAD) for drivers of SAE Level 2 automated vehicles. PCAD quantifies task difficulty using the gap between the current velocity and the safe velocity region in 2D, and accounts for the minimal control effort (braking and/or steering) needed to avoid a potential collision, based on visual looming, behavioural uncertainties of neighbouring vehicles, imprecise control of the subject vehicle, and collision severity. The PCAD model predicts both continuous-time perceived risk and peak perceived risk per event. We analyse model properties both theoretically and empirically with two unique datasets: Datasets Merging and Obstacle Avoidance. The PCAD model generally outperforms three state-of-the-art models in terms of model error, detection rate, and the ability to accurately capture the tendencies of human drivers’ perceived risk, albeit at the cost of longer computation time. Our findings reveal that perceived risk varies with the position, velocity, and acceleration of the subject and neighbouring vehicles, and is influenced by uncertainties in their velocities.

A large number of studies have proved that camera and radar fusion is a useful and economical solution for traffic object detection. However, how to improve the reliability and robustness of fusion methods is still a huge challenge. In this paper, an adaptive traffic object detection method based on a camera and radar radio frequency Network (CRRFNet) is proposed, to solve the problem of robust and reliable traffic object detection in noisy or abnormal scenes. Firstly, two different deep convolution modules are designed for extracting features from the camera and radar; Secondly, the camera and radar features are catenated, and a deconvolution module is built for upsampling; Thirdly, the heatmap module is used to compress redundant channels. Finally, the objects in the Field of View (FoV) are predicted by location-based Non-Maximum Suppression (L-NMS). In addition, a data scrambling technique is proposed to alleviate the problem of overfitting to a single sensor by the fusion method. The existing Washington University Camera Radar (CRUW) dataset is improved and a new dataset named Camera-Radar Nanjing University of Science and Technology Version 1.0 (CRNJUST-v1.0) is collected to verify the proposed method. Experiments show that CRRFNet can detect objects by using the information of radar and camera at the same time, which is far more accurate than a single sensor method. Combined with the proposed data scrambling technology, CRRFNet shows excellent robustness that can effectively detect objects in the case of interference or single sensor failure.

Crowded environments are inherently vulnerable to a range of risks, including earthquakes, fires, violent attacks, and terrorism. In such scenarios, every second counts in an evacuation, as it can significantly impact the number of lives saved. This paper introduces a novel approach to optimising crowd evacuation processes, focusing on behavioural modification rather than traditional methods such as mathematical optimisation models or architectural adjustments. We propose that by altering the behaviours of individuals within a crowd, overall system efficiency can be enhanced from within. We explore the effects of imparting simple, easily understandable strategies or instructions to individuals that can improve evacuation efficiency. The current work concentrates on how modifications in individual decision-making—namely, exit-choice and exit-choice-changing behaviour—can influence evacuation dynamics. We present the results of six major evacuation experiments, encompassing nearly 100 experimental scenarios and repetitions, which specifically investigate the effect of influencing exit choice and adaptation in exit-choice behaviour. The investigation revolves around three core questions: (a) the impact of effective strategies (b) the potential consequences of detrimental strategies, indicative of common misconceptions or poor advice, and (c) the influence of varying levels of strategy adoption, examining how system efficiency changes as more individuals embrace either beneficial or harmful strategies. The findings indicate that behavioural modification can substantially influence evacuation efficiency. Interestingly, the negative impact of poor strategies outweighs the benefits of effective ones. With respect to beneficial strategies, a significant increase in efficiency is observed at initial and intermediate levels of strategy adoption/uptake, suggesting that complete compliance is not necessary to enhance overall system performance. The benefit of influencing decision adaptation behaviour is considerably more tangible than influencing exit choice behaviour. These insights establish a novel perspective in evacuation safety. They lay a foundational framework for developing targeted public education and training programs based on empirical evidence. They highlight the importance of awareness and self-regulation among crowds, showcasing their potential to significantly increase both efficiency and safety in evacuation scenarios, potentially saving lives.

Electric Mobility-as-a-Service (E-MaaS) emerges as a promising solution for environmentally-friendly mobility in the future, yet MaaS operators have been struggling to achieve profitability. We introduce a novel E-MaaS ecosystem where platforms can leverage carbon credits revenue from the government’s emissions reduction fund (ERF) by incentivizing travelers to choose more E-MaaS services, thereby reducing carbon emissions. In such an E-MaaS ecosystem, travelers can select either electric (E)-MaaS or traditional (T)-MaaS services and submit heterogeneous requests, such as distance, service time, tolerance for inconvenience, and travel delay budget, which are modeled as inputs. We propose a multi-leader multi-follower game (MLMFG) model where each leader (MaaS platform) competes to maximize its profits by making operational decisions such as pricing, EV acquisition ratio, and E(T)-MaaS bundle allocation while anticipating travelers’ participation levels. In response to the platforms’ decisions, each follower (traveler) aims to minimize her travel costs by determining the participation levels for E(T)-MaaS services via multiple MaaS platforms. We develop a customized alternating direction method of multipliers (ADMM) algorithm to solve the proposed MLMFG efficiently. Comprehensive numerical experiments based on real-life data in Australia demonstrate the convergence and robustness of the proposed ADMM algorithm. Further, experimental results reveal how factors such as market size, travel demand, ERF budget, subsidy rate, and unit price boundaries impact the profits and operational strategies of different MaaS platforms. Overall, the proposed MLMFG model for the E-MaaS ecosystem provides valuable insights for MaaS operators aiming to balance profitability with environmental responsibility, navigating a future where sustainability and profitability goals could converge.

In the context of evacuating crowded spaces during acute crises, every second is pivotal and can be the determinant in life-or-death situations. It is, therefore, important to explore and implement any measures or interventions that could streamline and expedite the evacuation process in such scenarios. This study aims to explore how the modification of individual behaviours can be leveraged to improve the efficiency of crowd evacuations, with a specific focus on the physical aspects of movement. We examine three crucial elements of physical movement: behaviours at bottlenecks, the initiation time of individual movement, and the pace of movement. A series of dedicated experiments, each tailored to one of these behavioural aspects, has been conducted. In these experiments, the behaviour of interest is modified incrementally within the crowd, with increases of 20% at each stage. This methodology allows for a detailed assessment of system efficiency at varying levels of instructed behaviour adoption/injection. The findings reveal that changes in each aspect of physical movement significantly influence overall efficiency. Most notably, the relationship between the uptake and increase in efficiency is nearly linear, and the rate of efficiency increase does not notably diminish with uptake, unlike interventions pertaining decision-making aspects of behaviour. This suggests that behavioural interventions targeting physical aspects of movement will likely yield higher efficiency returns. Moreover, in comparison with a related study focusing on decision-making aspects of evacuation behaviour, this research observes that modifying physical aspects of behaviour is generally more straightforward. The success rates of individuals in implementing physical movement instructions are higher, and the impact on the system is more pronounced than that observed in decision-making modifications. These results provide insights for developing simple, actionable instructions that can be effectively communicated to the public. These instructions can be disseminated as part of training and education programs or even provided on the spot during an evacuation.

The generation of synthetic households is challenging due to the necessity of maintaining consistency between the two layers of interest: the household itself, and the individuals composing it. Hence, the problem is typically tackled in two steps, first focusing on the individual layer and then on the household layer. The existing two-step simulation method proposes generating the households based on their roles which diminishes the generality of the approach and makes it difficult to reproduce despite its beneficial properties. In this paper, we propose an alternative extension of Gibbs sampling for generating hierarchical datasets such as synthetic households, in order to make simulation more general and reusable. We demonstrate the performance of our method in a case study based on the 2015 Swiss micro-census data and compare it against state-of-the-art approaches. We show the influence of modeling decisions on different performance metrics and how the analyst can easily enforce consistency while avoiding generating illogical households. We show that the algorithm maintains the conditional distributions while satisfying the marginals of all variables simultaneously, all while generating consistent synthetic households.

This research introduces a novel approach to cooperative decision-making among self-organizing connected and autonomous vehicles (CAVs). In this approach, a coalitional game is played by a group of players who form alliances of different sizes based on the collective payoff they receive. The players continuously evaluate the potential benefits of different coalition formations and adjust their decisions accordingly. The proposed approach utilizes the V2V communication feature of CAVs, which enables CAVs to participate in a cooperative game, thereby resolving conflicting situations that often arise during lane-changing decisions. By working together within the same coalition, CAVs on a hypothetical three-lane freeway segment can collectively determine their target lanes, rather than engaging in individual decision-making that could result in a win-lose situation. The proposed approach considers up to nine CAVs interacting with each other and aims to find Pareto-optimal coalitions in lane-changing decisions. The approach considers lead CAVs that cooperate via acceleration to enlarge the gap between the subject and lead CAVs. The game is modelled as a dynamic transferable utility problem, allowing the utilities obtained from the coalition agreement to be expressed as real numbers and distributed among coalition members. The framework is generalizable to other traffic and demand management problems while the cooperative CAVs can be compensated for reaching an agreement in a universal, collectible, and tradable credit scheme (UCTCS) that can be used in a wide spectrum of traffic and demand management applications. The effects of the proposed coalitional lane-changing decision-making on traffic efficiency are compared to a non-cooperative decision-making model on a simulated road segment. Overall, our analysis suggests that the proposed coalitional approach can positively impact macroscopic traffic characteristics, leading to potentially improved traffic flow, reduced congestion, and enhanced travel time efficiency.

Decentralized traffic signal control methods, such as max-pressure (MP) control or back-pressure (BP) control, have gained increasing attention in recent years. MP control, in particular, boasts mathematically-proven network throughput properties, enabling it to optimize network throughput and stabilize vehicle queue lengths whenever possible. Urban traffic volume is dynamic and features a non-uniform distribution throughout the network. Specifically, heavier traffic is often observed along arterial corridors or major origin–destination streams, such as those in central business districts (CBD), while less traffic is found on sub-arterial roads. To address these issues, many existing signal plans incorporate coordinated signal timing. Numerous previous studies have formulated signal coordination optimization as mixed-integer programming problems, with most belonging to centralized traffic signal controller categories. However, centralized approaches do not scale well to larger city networks. In this paper, we introduce a novel max-pressure signal control approach called Smoothing-MP, which considers signal coordination in urban networks to achieve both maximum vehicle stability and reduced travel time and delay along specific urban corridors, without altering the original stable region proposed by Varaiya (2013). This study represents a pioneering effort in modifying max-pressure control to incorporate signal coordination. Crucially, this policy retains the decentralized characteristic of the original max-pressure control, relying exclusively on local information sourced from upstream and downstream intersections. To evaluate the proposed Smoothing-MP control, we executed simulation studies on two different types of networks, the Downtown Austin Network and a Grid Network. The results unequivocally show that Smoothing-MP matches the maximum throughput of the original MP control. Moreover, it significantly reduces both travel time and delay along coordinated corridors. This dual accomplishment underscores the efficacy and potential advantages of the Smoothing-MP control approach.

Deep neural networks (DNNs) have been increasingly applied in travel demand modeling because of their automatic feature learning, high predictive performance, and economic interpretability. Nevertheless, DNNs frequently present behaviorally irregular patterns, significantly limiting their practical potentials and theoretical validity in travel behavior modeling. This study proposes strong and weak behavioral regularities as novel metrics to evaluate the monotonicity of individual demand functions (known as the “law of demand”), and further designs a constrained optimization framework with six gradient regularizers to enhance DNNs’ behavioral regularity. The empirical benefits of this framework are illustrated by applying these regularizers to travel survey data from Chicago and London, which enables us to examine the trade-off between predictive power and behavioral regularity for large versus small sample scenarios and in-domain versus out-of-domain generalizations. The results demonstrate that, unlike models with strong behavioral foundations such as the multinomial logit, the benchmark DNNs cannot guarantee behavioral regularity. However, after applying gradient regularization, we increase DNNs’ behavioral regularity by around 6 percentage points while retaining their relatively high predictive power. In the small sample scenario, gradient regularization is more effective than in the large sample scenario, simultaneously improving behavioral regularity by about 20 percentage points and log-likelihood by around 1.7%. Compared with the in-domain generalization of DNNs, gradient regularization works more effectively in out-of-domain generalization: it drastically improves the behavioral regularity of poorly performing benchmark DNNs by around 65 percentage points, highlighting the criticality of behavioral regularization for improving model transferability and applications in forecasting. Moreover, the proposed optimization framework is applicable to other neural network–based choice models such as TasteNets. Future studies could use behavioral regularity as a metric along with log-likelihood, prediction accuracy, and $F_1$ score when evaluating travel demand models, and investigate other methods to further enhance behavioral regularity when adopting complex machine learning models.

With the rapid increase in the percentage of the world’s population living in cities, the design of existing transportation infrastructure requires serious consideration. Current road networks, especially in large cities, face acute pressures due to increased demand for vehicles, cyclists, and pedestrians. Although much attention has been given to improve traffic management and accommodate the increased demand via coordinating and optimizing traffic signals, research focused on adapting the static allocation of street spaces and right-of-way dynamically based on mixed traffic flow is still scarce. This paper proposes a multi-agent reinforcement learning (RL) agent approach that cooperatively adapts the individual lane widths and right-of-way access permissions based on real-world mixed traffic flow. In particular, multiple cooperative agents are trained with mixed temporal data that learn to decide suitable lane widths for motorized vehicles, bicycles, and pedestrians, along with whether co-sharing space between pedestrians and cyclists is safe. Using a microscopic traffic simulator model of a four-legged intersection, we trained our RL agent on synthetic data, and tested it on realistic multi-modal traffic data. The proposed approach reduces the overall average waiting time and queue length by 48.9% and 37.7%, respectively, compared to the Static (baseline) street design. Additionally, we observe CALM’s ability to gradually adjust lane widths, contrasting with the Heuristic implementation’s erratic lane adjustments, which pose potential safety concerns. Notably, the model learns to adaptively toggle the co-sharing of street space between cyclists and pedestrians as one co-shared lane, ensuring comfort and maintaining the level of service according to the designer’s policy. Finally, we demonstrate CALM’s scalability on a simulated large-scale traffic network.