Special issue: Smart transportation: Theory and practice

Transportation network, for providing mobility to all travelers, is an indispensable component of daily life in our society. With rapid urbanization and economic development, the continuously increasing travel demands in urban areas have resulted in serious traffic congestion problems in many large cities over the world. The adverse impacts of traffic congestion include not only considerably increased journey times and vehicular traffic delays but also fuel consumption and air pollution problems 1, 2. The conventional method for alleviating traffic congestion problems is to build more road infrastructures and/or expand the existing transportation networks. Yet, the use of the conventional method is increasingly restricted in most urban areas, because of the scarcity of land and public funds. To meet the challenges of providing adequate mobility in urban areas, smarter solutions are required by the best use of existing transportation facilities. In recent years, the advances of sensing and information technologies have produced a variety of spatiotemporal big data for travel in urban areas 3-6. The data sources include taxi trajectories, mobile phone records, smart card data, social media data, and various user-generated geographical information. These spatiotemporal big data for travel consist of the daily activity pattern and travel behaviors of travelers together with their interactions on transport-related environment. These extensive big datasets involve a large sample size of the total population. Such huge spatiotemporal data have offered a golden opportunity for developing advanced models and algorithms to improve transportation safety, enhance network efficiency and reliability, and reduce environmental impacts. The spatiotemporal big data together with advanced models and algorithms have a great potential to drive cities toward smart transportation. This special issue is devoted to the dissemination of the state-of-the-art research on the smart transportation topics ranging from theory to practice. In total, eight papers are included in this special issue and are briefly summarized in the succeeding text. Several of the papers included in this special issue present novel approaches to estimate and/or predict transportation network conditions using the emerging spatiotemporal big data that are collected from multiple sources. Neumann et al. 7 addressed the traffic data fusion problem of using multiple data sources to make better estimation of travel times instead of using a single data source. Based on Markowitz' portfolio theory in finance, they proposed an optimized weighted-mean approach to determine the optimal fusion of multiple data sources. The similarities and differences between the traffic data fusion problem and portfolio investment problem were discussed. The benefits of using negative weights and the ways of reducing systematic errors in the context of traffic data fusion were analyzed from the perspective of Markowitz' portfolio theory. Real-world floating car data of two independent vehicle fleets in Athens, Greece, were collected for case study of the proposed approach. The accuracy of the proposed optimized weighted-mean approach was examined with a comparison with a naïve fusion approach in which equal weights were used for all data sources. It was reported that the proposed approach can significantly reduce the systematic errors and variance of the fusion results, compared with the original single source of data and the naïve fusion approach. Further study was suggested to find better options for calibrating the systematic error correction factors and the covariances of data sources so as to improve the data fusion accuracy. Along a similar line of data-driven approaches, Wen and Yan 8 investigated the problem of predicting work zone traffic capacity for smarter work zone traffic managements. With consideration of capacity variation, they proposed a truncated lognormal model to predict the work zone capacity distribution. In the proposed model, the linear relationship was established between the work zone capacity distribution and nine decision factors, including road type, number of closed lanes, number of opened lanes, lane closure location, work duration, work intensity, work time, heavy vehicle percentage, and capacity measurement method. The maximum likelihood method was utilized to determine model parameters for predicting work zone capacity distribution. To validate the proposed model, 242 datasets from previous work zone projects in 14 USA states and cities were collected for a case study. It was reported that the proposed model can predict the range of work zone capacity with a satisfied accuracy. Results of case study found that work zones, located in urban roads with a large number of opened lanes, low heavy vehicle percentage or having long-term work duration, tend to have larger mean work zone capacity. The proposed model developed for freeway work zones can be extended for congested urban roads in further studies. The quality of spatiotemporal data has a direct influence on the accuracy and robust of data-driven approaches. With on-board GPS devices, Rim et al. 9 proposed a filtering-based model to detect and correct the outliers of GPS speed measures. In the proposed model, the locally weighted regression (LWR) filter was adopted. The mechanism for calibrating model parameters was discussed. A highway section in the Seoul, Korea, was selected to conduct field experiments. To validate the performance of the proposed LWR model, the exponential smoothing and autoregressive integrated moving average methods were also implemented for comparison of studies. It was reported that the proposed LWR model outperform the exponential smoothing and autoregressive integrated moving average methods in most cases. Further works can be conducted to detect and correct the outliers of speed measures in the floating car dataset with a low sampling frequency. Other papers in this special issue address various issues of developing smart controls of signalized intersections. Bie et al. 10 proposed an optimization model to improve the coordinated signal settings of hook-turn intersections. The proposed model was formulated as a mixed nonlinear integer problem for minimizing the average vehicle delays in the study network. In the proposed model, the average vehicle delay at intersections was calculated with explicit consideration of the spillback phenomenon of hook-turn vehicles in waiting areas. The platoon dispersion model was utilized to represent the traffic movements between coordinated intersections. Because of the complex nonlinear optimization problem involved in the proposed model, a genetic algorithm was developed. To demonstrate the effectiveness of the proposed model and solution algorithm, three intersections in downtown Melbourne, Australia, were selected for a case study in which the VISSIM simulator was used to generate vehicle movements under different signal settings. The results of the case study indicated that the optimized coordinated signal plan can significantly reduce the average vehicle delay and the number of spillbacks in both the peak hour and off-peak hour scenarios, compared with the current signal plan. Future work should be carried out to validate the proposed model and algorithm with empirical data collected at hook-turn intersections in practice. Kala 11 proposed an agent-based model to investigate the feasibility of developing cooperative transportation systems, which enables a few emergent vehicles to reach their destinations on time with a priority. In the proposed model, cooperative traffic lights were set to give a priority to emergent vehicles at intersections. A cooperative lane change mechanism was devised to make the vehicle changing lanes when an emergent vehicle is running behind it. To validate the proposed model, the road network of Reading, UK, was selected for case study. The simulation results showed that the cooperative transportation systems can significantly reduce the number of emergent vehicles reaching their destinations late, compared to that without cooperative mechanisms. The start-up times (SUTs) of the first two vehicles significantly affect the total start-up lost time at signalized intersections. Using collected video record data in Beijing, China, Li et al. 12 empirically investigated the relationship between the SUT of the first two vehicles and the total start-up lost time at signalized intersections. Three groups of data, including SUTs of the first two vehicles and car-following start-up time of the second vehicle, were extracted from the video records. Multiple linear regression technique and fuzzy C-means clustering technique were utilized for data analysis. The results of the data analysis showed that vehicle type and SUT of the first vehicle are major factors influencing the SUT of the second vehicle. Reducing SUT of the first vehicle is crucial to improve the total start-up lost time at signalized intersections. A last group of authors deal with approaches for making smart transportation policy and planning. Tahmasseby et al. 13 examined the potential of using a social network based carpooling system for staff and students of University of Calgary. A survey combined with revealed and stated preferences was conducted, and totally, 241 responses were collected for data analysis. A binomial logit model and two ordinal logit models were established to investigate the impacts of several socioeconomic, psychological, and travel characteristic factors on the carpooling participations. The survey results showed the existence of an unexploited potential for a social network based carpooling system at the University of Calgary. It was reported that low income commuters and students were more willing to use social network-based carpooling services. It was highlighted that the perceived rider and driver profiles in the social network can have significant impacts on the success of the carpooling system. Many other factors also have significant impacts on carpooling demand, including marital status, working schedule flexibility, trip characteristics, weather condition, and carpooling fee. Traffic assignment problem is an important issue for transportation planning. Ryu et al. 14 proposed a customized path-based algorithm for solving the stochastic user equilibrium (SUE) traffic assignment problem with multiple user classes. The proposed algorithm utilized a column generation scheme to generate the path set, an iterative balancing scheme to find the search direction, and a self-regulated averaging line search scheme to determine the step size. In the proposed algorithm, the vehicle restrictions in road networks and asymmetric interactions between different types of vehicles were explicitly considered. The route overlapping using the path-size logit model was adopted to take account random perceptions of network conditions in the SUE principle. To examine the computational performance of the proposed solution algorithm, a real network of Winnipeg city was adopted, and sensitivity analyses were conducted with respect to different model parameters. The results of computational experiments showed that the proposed solution algorithm is effective and robust to solve the multi-class SUE traffic assignment problems with different model parameter settings. While all these eight papers address different issues and present various approaches for improving theory and practice of smart transportation concept, they altogether highlight many challenges and opportunities for developing advanced models and algorithms using the emerging spatiotemporal big data in order to make various transportation systems smarter. The editors deeply wish that this special issue will serve as a kick-off departure for practicing engineers and researchers to conceive new perspectives, models, and solution algorithms for further development of new smart transportation systems in our society.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.