Journal of Transportation Engineering, Part A: Systems

Abstract

In this paper, we present a brief review of the most prevalent computer vision–based traffic signal recognition studies in the literature. Based on the adopted computer vision approaches, we classify the traffic signal recognition studies into three categories: model-based, classical machine learning–based, and deep learning–based methods. Additionally, we include an extensive analysis of the traffic signal data sets used for training and evaluating traffic signal recognition deep learning models. This paper provides researchers and practitioners with insight into the research trends in traffic signal recognition used in vehicle perception, emphasizing various adopted methodologies and their detailed performance parameters.

Practical Applications

Traffic signal recognition is a crucial component of autonomous driving and involves detecting and recognizing different states of traffic signals. Over the years, researchers have explored a wide range of digital image processing and machine learning methods for traffic signal recognition applications. This paper provides a brief literature review of the most prominent traffic signal recognition studies. Innovations in this field are crucial to revolutionizing how autonomous vehicles interact with the existing traffic signal infrastructure and the surrounding traffic, leading to improved safety and better traffic management. Some of the significant challenges in this field derive from the requirements of addressing different lighting conditions, weather, and occlusions, which makes the ongoing research indispensable.

Due to the effects of low-pressure and hypoxic environments at high altitudes, drivers in high-altitude areas exhibit increased perceived reaction times, leading to challenges in accurate speed estimation and handling judgment. This study aims to quantitatively analyze the impact of plateau environments on operating speeds at interchange exit ramps. Utilizing a UC/win-road simulator, six scenarios of expressway exit ramps were constructed. The simulation experiments involved 50 participants (35 males, 15 females) from Nanjing, China (altitude of 50 m) and 50 participants (36 males, 14 females) from Lhasa, China (altitude of 3,650 m). This research focused on examining the influence of the plateau environment on drivers’ operating speeds, investigating variations in speed between drivers in plain and plateau areas, across genders, and during different acclimation periods. It also aimed to predict operating speeds at the midpoint and exit of the curve on the exit ramp for drivers in both plain and plateau areas. Based on these predictions, the study elucidated the trend of operating speed as influenced by the low-pressure and hypoxic conditions of the plateau, as well as the characteristics of the exit ramp’s horizontal curve. Additionally, the research uncovered the internal correlations and potential reasons linking operating speed to drivers’ perception and response abilities, physiological and visual load levels, and driving styles.

Automated vehicles equipped with adaptive cruise control systems (ACC-AVs) are prevalent and the ensuing issue is the feasibility of ACC-AVs’ operation on the roads. This study investigates the feasibility of road horizontal curve designs for ACC-AVs from a vehicle dynamics perspective. Following the scenario generation framework, we created and tested several scenarios featuring horizontal geometric elements and design speeds, conducting a safety evaluation based on the critical adhesion coefficient, lateral acceleration, lateral-load transfer rate, together with driving comfort indicators. Results indicate that ACC-AV can navigate on road curves designed with a common minimum radius ($R_{\min \text{\_}com}$) effectively at speeds over $60\mathrm{km}/\mathrm{h}$, comparable to conventional vehicles. However, both $R_{\min \text{\_}com}$ and limited minimum radius ($R_{\min \text{\_}\lim }$) designs show limitations. Additionally, the feasible radius ranges for ACC-AV reveal the capability to safely handle sharper curves and maintain higher speeds, suggesting potential for adaptable road design in complex environments. Finally, minimum radius ranges were summarized for ACC-AV safe and comfortable operation on road curves, unveiling the potential risks and reminding designers in curve design controls for ACC-AVs.

Focusing on the insufficient passenger demand during off-peak periods in urban rail transit, a new train operation mode with flexible train composition considering trains couple/decouple at intermediate stations is proposed. An optimized two-phase nonlinear model is formulated in order to design a periodic train schedule, where train headways remain constant over a specific period. This model takes into account the flexible train composition. The objective function aims to minimize total passenger waiting time and train operating costs while considering constraints such as maximum line passing capacity, passenger flow demand, and the number of train units in the composition. Linearization techniques are applied to convert nonlinear constraints into linear equivalents, and the CPLEX solver is employed to solve the model. An illustrative real-case example is presented to validate the effectiveness of the proposed model. The results demonstrate that the operation mode with flexible train composition during off-peak periods offers significant advantages over traditional train composition modes, including single train routes, long-short train routes, and multiple train compositions. This mode enables better alignment between passenger demand and capacity, reduces operating costs, and enhances passenger service levels. The proposed operation mode is particularly suitable for urban rail transit lines with low passenger demand but requiring high passenger service, such as suburban lines or common lines during off-peak periods.

This study focuses on solving the problem of metro first/last mile, studying the method for designing microcirculation bus routes and coordinating schedules considering the impact of shared bicycles. First, we propose a bilevel mixed-integer programming model for designing microcirculation bus routes and coordinating schedules considering the impact of shared bicycles. The upper-level model minimizes the weighted sum of the travel time cost of passengers and the operating cost of public transport enterprises, and the lower-level model maximizes the number of passengers served by microcirculation bus routes. Then, an improved genetic algorithm is developed to solve the model, called the Monte Carlo adaptive genetic algorithm (M-GAI). Finally, the proposed model and algorithm are evaluated using the case study in the area near the Fubao metro station of Shenzhen Metro Line 3. Results show that if the impact of shared bicycles is not considered, the passenger demand will be greater than the actual value, and the operating cost of public transport enterprises will be increased by 36%. Compared with GAI, the average number of iterations of M-GAI is reduced by 31%, and the objective function value is decreased by 4%. In addition, when the number of routes increases, the average waiting time of passengers is shortened, the average attendance rate of microcirculation buses increases, and the average empty distance of each vehicle is shortened. However, the operating cost of public transport enterprises will increase with the number of routes. Finally, when weight factors $\alpha$ and $\beta$ are 0.6 and 0.4, respectively, and the sum of the travel time cost of passengers and the operating cost of public transport enterprises reach optimal.

The Green Light Optimal Speed Advisory (GLOSA) system suggests speeds to vehicles to assist them pass intersections during green intervals. In application, drivers can ultimately decide whether to change speed based on their driving experience. However, with the gradual popularization of autonomous driving, drivers are gradually leaving the driving area. Therefore, the central algorithms in on-board systems need to be trained more intelligently for autonomous decision-making. Specifically, we found that the frequency of decision making can significantly affect the performance of the GLOSA system, but this issue has not been discussed in previous research. In this paper, we propose an adaptive frequency GLOSA (AF-GLOSA) model based on deep reinforcement learning (DRL) algorithm. Different from traditional models, this model can extract effective features from raw data and learn decision-making experience through constant interaction with simulated environments. By using parameterized action spaces, we divided the GLOSA task into two parts: frequency control and speed consultation. The frequency control module helps filter out unnecessary operations, and the speed consultation module provides acceleration suggestions based on the results of the upper model. In addition, we have designed a novel reward function to balance fuel consumption and travel efficiency. Finally, the AF-GLOSA model was evaluated in both single intersection and multi-intersection scenarios in SUMO. The results indicate that the model can effectively reduce fuel consumption and carbon dioxide emissions in both cases. In term of the number of stops, the single-intersection outperforms the state-of-the-art method and the multi-intersection approaches the state-of-the-art method. The final results also demonstrate the necessity of considering decision-making frequency.

Frequent highway accidents occur in the Guizhou region, among which poor visibility due to fog is one of the main causative factors. In this region, traditional large-scale, high-computational-power fog monitoring systems are difficult to install and have high costs due to complex terrains, high altitudes, and winding roads, causing traffic management departments to fail to obtain fog information accurately and timely, which undoubtedly becomes a significant safety hazard. To solve this problem, this study proposes a fog monitoring solution based on the lightweight deep learning model ABNet. The solution first preprocesses the input images, including generating the fog concentration distribution map using the fog imaging model, and obtaining the high-frequency component image using filters based on the 2D discrete wavelet transform. Subsequently, these two processed images and the original image are fed into the three branches of the ABNet for training to fully extract fog concentration and high frequency information, thereby improving model performance and prediction accuracy. The ABNet model parameters only require 38.52MB, and the computational complexity is a mere 1.71GFLOPs, effectively solving the limited storage and computational resources problem in edge computing. The model was evaluated using the Guizhou highway fog weather data set, and ABNet exhibited impressive performance with a composite classification accuracy as high as 92.3%, reaching 92.4% in average precision rate and 92.3% in average recall rate. In comparison, the performance of models like VisNet, VGG16, EfficientNetV2, and Swin Transformer V2 seemed inferior. The experimental results validated the excellent performance of the ABNet model in terms of accuracy and efficiency. The ABNet model in this study, with its lightweight deep learning design, small parameter scale, and lower computational power requirements, provides a solution suitable for complex terrains and practical environments of edge computing devices, and it provides vital technological support to improve traffic safety on the highways in the Guizhou region.

The increased need for active transportation facilities coupled with the limited funding and space have influenced the prioritizing of shared-use paths (SUPs). Unlike other activity-specific facilities, the SUP can accommodate a wide range of users. With SUPs being relatively new facilities, less is known about the characteristics of the users and the key factors associated with the user type. This study explored the influential factors for SUP user diverse activities using multinomial regression on the survey data collected in Edmonton in 2018. The descriptive analysis revealed that walking was the activity with the highest frequency, followed by walking and cycling, and walking with pets, whereas cycling had the lowest priority. The multinomial model showed that as the age increases, residents are less likely to perform activities other than walking or cycling alone. Further, residents with higher education are more likely to either walk and cycle or walk, run, and cycle. Residents whose secondary mode of transportation is bicycle are less likely to walk and walk pets. Residents who own their house are likely to walk and walk pets. Furthermore, male residents, residents with children and those whose primary mode of transportation is not personal vehicles are more likely to walk, run, and cycle but less likely to walk and walk pets, compared with either walking or cycling alone. Planners can utilize the findings to understand the possible utilization of the planned SUPs and design them accordingly.

Accurately predicting train running times (TRTs) during disturbances is crucial for effective timetable rescheduling. Previous studies on train running prediction have overlooked the detailed textual description of disturbance events. To improve the prediction accuracy, this paper proposes a hybrid neural network model consisting of a transformer encoder and a fully connected neural network (FCNN) to predict the TRTs under disturbances, which is called transformer-FCNN here. In the proposed transformer-FCNN architecture, a transformer encoder is used to process textual disturbance events, while FCNN is used to deal with static features. The performance of transformer-FCNN is validated using data from the Wuhan-Guangzhou and Xiamen-Shenzhen high-speed railways. The results show that the models considering the disturbance category can substantially improve the predictive performance compared with those that do not. Further, based on the same algorithm, models which use textual disturbance event records are shown to reduce the mean absolute error and the root mean squared error, respectively, by over 10.8% and 8.3% on average compared with those that only consider the disturbance category. The proposed model can support potential strategies of train operation control, by providing accurate prediction results.

An electric carsharing system (ECS) is a crucial component in the development of a sustainable transportation system. Understanding user behavior toward ECS is essential for successfully implementing such a service. This study investigates the impact of behavioral intentions to use ECS and its determinants on traveler mode choice behavior. A hierarchical hybrid choice model is constructed to analyze the relationship between behavioral intentions and the willingness to choose ECS. The study collected data from 771 respondents in Dalian City, China. The results highlight the importance of behavioral intentions and their constructs, such as performance expectancy, effort expectancy, hedonic motivation, and familiarity with the ECS concept in influencing individuals’ willingness to choose ECS. Additionally, ECS attributes and social characteristics also significantly affect individuals’ decisions. These findings provide valuable insights for researchers and policymakers to promote ECS effectively.

Annual average daily traffic (AADT) is one of the most inevitable elements for both transportation planning and traffic safety analysis. For state Departments of Transportations (DOTs), collecting AADT data is a critical and demanding task, normally accomplished through a combination of permanent and temporary traffic count stations, which has been proved to be extremely labor-intensive and time-consuming. Consequently, due to limited resources, it is typically performed for the state-maintained highways rather than the low-volume roadways maintained by town jurisdictions. Therefore, it is necessary to develop innovative and cost-effective approaches to generate AADT for low-volume roadways while still maintaining accuracy, including data driven analytical methodologies. To this end, this study conducted a comprehensive literature review of methodologies and relevant data used for AADT generation and estimation. Based on the data availability, data elements used for modeling AADT in Connecticut (CT) were collected. Specifically, AADT data for town-maintained highways from 2016 to 2020 was collected from the Streetlight platform and calibrated to fit the local condition in CT. Furthermore, machine learning algorithms were developed to predict future AADT beyond 2020. The model validation results indicate that the AADT estimated by this study is robust and reliable in terms of prediction accuracy, and it can serve as a valid asset in transportation planning and traffic safety analysis to practitioners and transportation agencies.

Autonomous truck platoons hold the potential for substantial economic and environmental advantages. However, there is a lack of comprehensive research on the interactions between drivers and different configurations of truck platoons at waving segments. This study aims to contribute to the literature by achieving three primary objectives: (1) examine the effects of various truck platoon configurations on merging and diverging behaviors [time to merge (TTM) and time to diverge (TTD)]; (2) explore the impact of individual characteristics such as age, gender, education level, and driving experience on TTM and TTD; and (3) investigate the decision-making associated with these maneuvers (merging or diverging in front of the platoon, behind the platoon or cut in through the platoon). A driving simulator study was conducted with 85 participants across 12 distinct scenarios, considering variations in platoon size, intraplatoon spacing, and platoon lane-change behavior. Several statistical methods were employed, including ANOVA, the Cox proportional hazards model, and machine-learning techniques, to analyze the factors impacting TTM and TTD. The results revealed that increasing the headway distances between trucks in a platoon to 13.72 m (45 ft) or 18.29 m (60 ft) substantially decreased TTM, enhancing traffic flow. Furthermore, splitting the truck platoon between two lanes of the freeway before the merging point significantly influenced other drivers’ merging decisions. When half of the trucks in a platoon switched to the left lane before the merging point, a larger proportion of participants chose to merge ahead of the platoon. Age, gender, education level, and self-assessment of driving skills were all found to significantly influence merging and diverging behaviors. Drivers with higher degrees took longer to merge, whereas older, male, and experienced drivers merged faster. The lowest average TTD was observed when half the platoon switched to the left lane before the diverging point.