Drivers' route choice model is essential in transportation software such as navigation, fleet management, and simulation, where the random utility models (RUM) have dominated for years. The authors investigate here whether machine learning (ML) models could be applied into this field, and whether these approaches outperform the traditional models in goodness-of-fit and prediction. The application framework and data structure are proposed, where the challenging problems lie in: (i) to pool data from multiple origin–destination pairs; and (ii) to interpret results for behaviour analysis. All RUM and ML models are applied in a real network. Results suggest that the random forest, one of the ML models, has satisfying performances with acceptable computation time, making it suitable for large network and real-time analysis. This study shows that the ML models can be adopted for behaviour analysis, such as to prioritise the importance of variables, compute the elasticity, and forecast for scenarios. Future directions on combining the RUM and ML models are discussed.

With bus global positioning system data and smart-card data, this study puts forward improved algorithms to further increase the success rate of alighting stop identification for the student group, so that trip chains and origin-destination (OD) matrixes of students can be obtained with the high estimated rate. Aimed at the student group, after using and generalising conventional alighting algorithms, this study innovatively utilises resident and non-resident students’ typical trip patterns to further increase the success estimated rate. As a result, the algorithms are verified with a correct estimation of 74.9%, where the success rate steeply increases by 8.6% through the method based on students’ typical trip patterns. The empirical analysis and application have shown that the methodology can observe the trip chains of students and optimise the bus service for students. In the future, OD matrixes obtained should be validated with bus OD surveys, and traffic system of metro and shared bikes as well as other datasets such as land use data will be taken into consideration.

Travel decision making is driven by different activities. To better understand the traffic-activity characteristics of different regions, representative activity patterns (RAPs) of six Sino-American metropolises are recognised, and then these metropolises are clustered for each kind of RAP. Same year's National Household Travel Survey data of USA and Shanghai Household Travel Survey (SHTS) data are used in this study. Each resident's activity-topic probability distribution could be obtained by latent Dirichlet allocation topic model, then residents’ RAPs of a region are recognised by affinity propagation clustering, for each region's weekdays and weekends, respectively. Based on RAPs recognition, the six metropolises are clustered within conspecific RAP according to the dissimilarity of pattern characteristics (the number of per capita activities, sex ratio etc.). At last, six metropolises are divided into three, two, five clusters in accordance with three kinds of work patterns, which are weekdays’ normal work pattern, weekdays’ late return (home) work pattern, and weekends’ work pattern. Regions in the same cluster have homologous demographic compositions and show similar activity characteristics.

Most researchers and insurers evaluate driver's risks solely based on user global positioning system (GPS) trajectories. A new approach is proposed that collects individualised driving behaviour data from smartphone GPS module, combined with geographical network information and dynamic traffic conditions, to identify driving risk factors and evaluate driving behaviours in various contexts. The multi-source data reveals real world activity patterns as to when, where and how an individual driver performs, from which performance measurements on both the trip and driver level are defined to measure the risks. In addition, the relationship between the defined driving performance measurements and accident rate can be examined and verified by combining the crash history of the drivers. Numeric analysis on the trip level demonstrates that driving behaviour is context-sensitive, and the principal component analysis performed at the driver level shows that pedal operation and driving speed are the two most important performance measurements to characterise an individual's driving pattern. The subsequent correlation analysis of crash history and driving performances verify both pedal operation and driving speed are significantly related to more at-fault accidents, which validates the modelling and analysis efforts. The findings of their study further existing knowledge and provide foundations for advanced pay-as-you-drive-and-you-save insurance pricing.

Link travel time is an important evaluation index of urban arterial traffic condition. It is traditionally estimated from loop detector data and GPS probe vehicle data. The former suffers from the coverage and maintenance issues. The latter has issues related to low penetration rate. Cellphone handoff data is one alternative that may address those deficiencies with its large coverage areas and continuous detection capabilities at low cost. Existing cellular probe vehicle based travel time estimation methods are primarily designed for freeway. Limited studies have been conducted for signalised arterials. Cellular handoff data detects the time lapse between two consecutive wireless handoff points. When vehicles with on-call cellphones traverse between two handoff points, their space mean speeds can be calculated by dividing the pre-calibrated distances by the time lapses. In this study, a new cellular handoff data based arterial travel time estimation model is proposed and compared with existing models. A hybrid traffic-and-wireless simulation network is built to simulate the handoff data generation and processing. Field data are collected at an arterial corridor in Chengdu, China to establish the baseline condition. The results from the proposed model show promising performance with detection errors of link travel times mostly within 10%.

Bicycling is a potential way to improve sustainable urban transportation system. However, the evaluation methods and metrics for bicycle ride quality and slow traffic pavement condition are still in the preliminary stages. The traditional bicycling vibration measurement used instrumented probe bicycle (IPB), which contains several independent devices and obtains scattering data. This method cannot show the detailed pavement condition. A technical challenge for monitoring real-time bicycling comfort and obtain the pavement condition in detail is to combine the intelligent transportation system (ITS) with bicycles. In this study, two schemes have been proposed to acquire the vibration data based on portable terminal measurement systems. Scheme (1) was based on the IPB using GPS, accelerometers, and sport cameras. The microcontroller was used to integrate several modules to develop scheme (2). Wavelet analysis was introduced to analyse random signal data. Then the pavement condition can be reflected in detail by vibration response. Data derived from the proposed system can be successfully applied to promote the use of bicycles as a reflecting tool for the pavement surface conditions. This method can provide a reference database for pavements’ surface maintenance and for cyclists to choose cycling routes.

Pavement Management System (PMS) is used to conduct efficient pavement management. Since the pavement network is generally large and complex in scale, the reliability of the pavement network cannot be easily determined by applying conventional serial or parallel reliability models, or even a combination of the two. The objective of this paper is to propose a novel model by employing fuzzy set theory and dual-network method to calculate the relative importance of each pavement route and estimate the reliability of a pavement network. A fuzzy condition model is developed to describe the condition state of pavement management segments, pavement links, pavement routes, and eventually the whole network. The spatial weight of each pavement route is represented by the node degree of the corresponding dual-network. A case study is conducted by using the pavement network of Travis County, Texas, to illustrate the applicability of the proposed methodology. The results show that the model is able to calculate the relative importance of each route and determine the reliability of the given pavement network. This paper could help transportation agencies obtain more insight information on relative importance of each roadway and make more effective decisions on pavement network maintenance with a limited budget.

Modern roundabouts are widely used at intersections with light traffic, generally providing safety and other advantages. However, large entry delays are often observed at roundabouts with unbalanced flow patterns, even though the entry traffic flow is not high. A metering signal-based strategy is examined to mitigate the above problems. A mathematical optimisation model is formulated firstly with the objective of minimising the total entry delay, subject to the metering signal thresholds. Then a solution algorithm based on VISSIM simulation is developed. Finally, a case study is carried out to testify the feasibility and applicability of the proposed model. Extended scenarios analyses under different levels of approach volume, different demand combinations and different proportions of right-turn vehicles (left-side driving) are also conducted. Results show that the methodology can effectively improve the operational performance, and a delay reduction of up to 25.7% can be expected using the metering signal-based strategy. This can provide a criterion for the use of metering at the roundabout.

For enhancing the intelligence of urban traffic, connected automated vehicle (CAV) is recognised as the leading technology in the near future. With on-board sensors and communication devices, status of the vehicles can be obtained to better coordinate the traffic. However, the limited environmental perception range cannot lead to the best efficiency of the global urban traffic. In this study, a three-step evolution strategy of the CAV perception mode is proposed, from autonomous perception to interactive perception to networked perception. Key technologies in these three steps are studied. In autonomous perception, vehicle positioning and dynamic target tracking approaches are proposed. In interactive perception, a reliable multi-mode information exchange mechanism is studied. Finally, a new traffic big data storage and advanced analytics solution in networked perception is introduced. Related experiments about the above key issues are designed, implemented, and verified based on the hardware platform, open source dataset, and cloud platform, respectively. The results show that the positioning distance root mean square achieves 3.9 m, object tracking speed reaches 30fps, and communication average packet loss rate is 2%. As can be seen from testing and simulation results, our proposed approaches can meet technical requirements and support environment perception mode evolution of CAV.

Mandatory lane change (MLC) is a critical step in formulating global route of an autonomous vehicle on the urban road network. Improper MLC decisions on arterial roads could jeopardise efficiency (travel cost) and reliability (possibility of failed lane change). However, the existing research studies seldom investigate the optimal MLC decision at the planning level to maximise the reliability-based driving efficiency. This research aims at addressing two core strategic decision variables (MLC decision point on the road and maximum waiting time before giving up MLC). A series of simulation experiments for various scenarios are conducted for an arterial road to reveal their relationship. The results indicate that both identified decision variables inherently affect travel time spent on this road and the rate of failed MLCs, and a trade-off exists between arterial travel time and the rate of failed MLCs. Based on the simulation analysis, an analytical lane-level link performance (LLP) function is formulated to assess the impacts of MLC decisions on driving efficiency. The analysis validates that the optimal MLC strategic decision (MLC decision position and maximum waiting time) can be determined by maximising LLP function. It is promising to apply the proposed LLP function in lane-level routing algorithm in the future.

A nighttime vehicle detection method in the multi-lane intersection has been proposed based on saliency detection for traffic surveillance system in this study. Frame difference method is applied to detect moving objects at first, and all the rear lights of vehicles are extracted based on saliency map and colour information. Second, vehicles are detected through pairing off all the lamps, which include such steps as rechecking tail-lamp pairs by using prior knowledge, eliminating repaired tail-lamps on the same vehicle and removing the paired lamps across two lanes. Furthermore, to detect the vehicles that only have a single valid tail-lamp, a proving approach for virtual tail-lamp is investigated. Finally, the proposed method is verified to be more reliable and faster for nighttime vehicle detection by comparing with other detection methods, which can satisfy real-time requirements of a vehicle detection system with good performance.

Bus running times are a key element of time reliability for bus operators and passengers. Hence, their evaluation is crucial in order to build reliable schedules for transit operations. This study analyses three alternative offline methods for estimating bus running times using automatic vehicle location (AVL) data: the average method, the percentile method and an adjusted Kalman filter method, which is amended in order to be implemented for offline use. Experiments are conducted using ∼92,000 real-world archived AVL records, which are provided by an Italian bus operator. The results can be used to revise scheduled running times along a specific route by these methods.

The location of the U-turn median opening at signalised intersections can have important effects on the intersection performance. However, there has been very little prior work on this issue, especially on intersections with double left-turn lanes. This study analyses the mutual interference of the U-turn and the left-turn movements at signalised intersections with double left-turn lanes. A lane selection model, based on the Wardrop Equilibrium Principles is developed for left-turn drivers. The saturation flow rate model of the double left-turn lanes is developed to quantitatively evaluate its utilisation. This study also presents the guidelines for the design of the U-turn median opening for maximising the saturation flow rate of double left-turn lanes. The proposed model is also verified by means of simulation, and then a case study is used to compare the intersection performance before and after optimisation. The results of case study show that if the U-turn locations are varied, the saturation flow rate of the double left-turn lanes would also vary considerably. The authors show that there can be a major difference of 77.47%, and the saturation flow rate of optimal location can be improved by 7.63% compared with the current location of the U-turn median opening.

The collaboration between a human driver and an automation system will serve as an effective measure before the autonomous driving technology is fully implemented. To discuss the driving authority between a human driver and an automation system, a novel moving horizon shared steering framework is proposed, in which the controller assists the driver when the vehicle is in danger. First, a potential-hazard analysis is presented based on the potential steering operation error and lane departure distance to predict the degree of a hazard and to determine the driving authority between the human driver and the automation system. Then, the minimum collaborative steering operation is determined using a moving horizon optimisation approach with safety constraints. Thus, the automation system shares steering with the driver protect the vehicle from risks in a non-invasive manner. To demonstrate the effectiveness of the proposed strategy, simulations with different types of drivers and scenarios are conducted. The results of these simulations demonstrate that the proposed approach can enhance the driving ability of drivers with different skill levels in dangerous situations. The approach can also guarantee the safety of the intelligent vehicle to some extent when drivers lose focus for a period of time.

The limited availability of vehicle state information, including tire-road forces and vehicle velocities, restricts the development of control strategies for intelligent vehicles and new energy vehicles. This study proposes a modular estimation scheme for tire-road forces and vehicle velocities that can effectively cope with the cyclic coupling of the vehicle dynamics. The longitudinal tire-road forces are estimated using a sliding mode observer. Then, an observer for the lateral tire-road forces that exist in a cascade structure with the longitudinal tire-road force observer is designed. Non-linear vehicle velocity observers that take the estimated longitudinal and lateral tire-road forces as inputs are designed. A genetic algorithm approach is employed to select the observer gains. Finally, experimental validations under normal conditions and offline simulations under critical conditions for verifying the robustness with respect to measurement noise are conducted. The results demonstrate that the proposed modular scheme for tire-road force and vehicle velocity estimation yields acceptable results and has potential value for real vehicle applications.