Archives Search

Abstract

Automated Driving Systems are migrating from R&D to market deployment. However, at the same time, the advances in electric propulsion are lowering the complexity of the Powertrain design. Furthermore, limiting and controlling the operating environment can significantly re...Lead moreAutomated Driving Systems are migrating from R&D to market deployment. However, at the same time, the advances in electric propulsion are lowering the complexity of the Powertrain design. Furthermore, limiting and controlling the operating environment can significantly reduce the requirements for the Automated Driving System. This combination is enabling new entrants to develop and introduce novel use-case specific automated driving vehicle designs for city applications. The success or failure for deployment will highly depend on the capabilities required to develop, integrate, verify and validate the self-driving technology into the vehicle, into the smart city eco-system (cloud), and into the digital and physical road infrastructure. This paper proposes a structured approach to define the engineering development and support functions required to develop and deploy such self-driving systems in an urban environment. The author has been working on ADS in Silicon-Valley, Europe and Asia and is currently based in Singapore.

Abstract

Intellectualization is an important characteristic with the high-quality development of transportation, which helps maximize the efficiency of transport infrastructure and improve the efficiency and service quality of the transportation systems. The development and applic...Lead moreIntellectualization is an important characteristic with the high-quality development of transportation, which helps maximize the efficiency of transport infrastructure and improve the efficiency and service quality of the transportation systems. The development and application of intelligent transportation technology could not be separated from the support and guidance of standardization. Based on summarizing the problems and deficiencies of the current situation of intelligent transportation standardization, combined with the development trend of intelligent transportation technology and the demand of superb development of transportation, we put forward the countermeasure proposal of intelligent traffic standardization development in this paper, which means to promote the development of intelligent transportation standardization and bring better service for the implementation of national strategy and the construction of traffic power in China.

Abstract

Shared parking is an effective way to alleviate parking difficulty and increase the utilization of parking facilities. This paper proposes a new allocation and pricing model for shared parking lots (SPLs). First, the overflow parking demand of the destination parking lot ...Lead moreShared parking is an effective way to alleviate parking difficulty and increase the utilization of parking facilities. This paper proposes a new allocation and pricing model for shared parking lots (SPLs). First, the overflow parking demand of the destination parking lot (DPL) is allocated to those vacant SPLs, with the aim of User Equilibrium (the time cost for each SPL is the same). The time cost consists of the search time for a parking space inside the SPL and the walking time from the SPL to the destination. The optimal allocation result is solved by the Method of Successive Average. Second, to encourage drivers to comply with the allocation, different parking prices of the DPL and SPLs are calculated with the method of elastic coefficient. Finally, a case study with Shanghai field data is analyzed to prove the feasibility of the model.

Abstract

This paper mainly introduces Intelligent Vehicle-Infrastructure Cooperation System based on 3D LiDAR technology and V2X communication technology as well as the roles of the system in creating a safe, efficient and intelligent "human-vehicle-road" ecosystem. Composed of ro...Lead moreThis paper mainly introduces Intelligent Vehicle-Infrastructure Cooperation System based on 3D LiDAR technology and V2X communication technology as well as the roles of the system in creating a safe, efficient and intelligent "human-vehicle-road" ecosystem. Composed of roadside LiDAR, V2X roadside unit, V2X on-board unit, on-board LiDAR and intelligent terminal equipment (such as smart phone), Intelligent Vehicle-Infrastructure Cooperation System subverts conventional V2P and V2I human-vehicle-road interaction mode. On-board 3D LiDAR and roadside 3D LiDAR complete the collection of road information with no blind area, and can realize the 4D reconstruction of road environment based on point cloud data. After the point cloud data feature extraction, the road traffic participant information can be sent to roadside V2X unit. V2X roadside unit encapsulates the extracted traffic participant information in a standard RSM message format and sends it out. On-board V2X unit decodes the received roadside V2X data, and extracts vehicle travel-related information. In brief, Intelligent Vehicle-Infrastructure Cooperation System can effectively eliminate the blind areas of traffic participants through the fusion of environmental perception data, and enable traffic participants to learn the traffic conditions on the driving path in advance, so as to effectively improve the traffic safety and the traffic operation efficiency.

Abstract

In this study, machine learning techniques and heuristic method are applied to long-term travel time prediction. We set up several criteria to evaluate different method for different scenario, trying to find the best combination of three methods for each target road segme...Lead moreIn this study, machine learning techniques and heuristic method are applied to long-term travel time prediction. We set up several criteria to evaluate different method for different scenario, trying to find the best combination of three methods for each target road segment. In practical experimental environments, the travel time records of Freeway No. 1, No. 3 and No. 5 were collected and used to evaluate the proposed method. The results showed that the mean absolute percentage error of the proposed is below 10% among all scenarios. Therefore, the proposed method based on machine learning techniques and heuristic method can provide precise long-term travel time prediction information for our system.

Abstract

Based on the research status of dynamic path system, a multi-objective path generation method for dynamic re-planning is proposed.It is mainly divided into two aspects: path selection model and path optimization algorithm. A multi-objective path selection model based on t...Lead moreBased on the research status of dynamic path system, a multi-objective path generation method for dynamic re-planning is proposed.It is mainly divided into two aspects: path selection model and path optimization algorithm. A multi-objective path selection model based on the shortest time, the shortest distance and the lowest congestion is proposed. The path solving algorithm is determined, the competitive learning neural network is improved to determine the congestion degree classification, and the global path planning is performed by the reverse A* algorithm, and the initial optimal path is given. When the information in the road network is detected to change, the new information is fed back into the system, and the dynamic update is performed by the incremental update algorithm to achieve real-time dynamic path planning, and the model algorithm is implemented by MATLAB simulation.

Abstract

Abstract: Measurements of the blood pressures at every 10 min. of an elderly driver suffering the hypertension during 30 km car driving showed that the systolic (maximal) blood pressure Pmax and the diastolic (minimal) blood pressure Pmin both were brought down with 28 mm...Lead moreAbstract: Measurements of the blood pressures at every 10 min. of an elderly driver suffering the hypertension during 30 km car driving showed that the systolic (maximal) blood pressure Pmax and the diastolic (minimal) blood pressure Pmin both were brought down with 28 mm Hg and 12 mm Hg, respectively in the averaged values at the magnetic stimulation applying 1-20 Hz, 100 mG magnitude, 4 ms width pulse train magnetic field to his spine position. Similar blood pressure decreasing effect was also obtained at another magnetic stimulation by setting a magnetized slag crashed stone aggregate at the spine position. These magnetic stimulation methods were predicted on the basis of so-called Magneto-Protonics principle, in which the free protons are activated in an electrolyte solution of the bio cells for promoting the ATP production rate. These magnetic stimulation effect is verified with measurements of the bio-magnetic signal at the left scapula bottom position using a pico-Tesla resolution magneto-impedance sensor and also decreasing the hysteresis of an electric conductivity push-pull temperature characteristics 20 to 0 to 20 deg C in the electrolyte solution.

Abstract

Mobility as a Service (MaaS) is the integration of various forms of transport service into a single mobility service accessible on demand. In order to improve the attractiveness of MaaS service, especially public transport services, based on the rapid flourish development...Lead moreMobility as a Service (MaaS) is the integration of various forms of transport service into a single mobility service accessible on demand. In order to improve the attractiveness of MaaS service, especially public transport services, based on the rapid flourish development of mobile internet plus with all kinds of consuming service, such as eating out, travelling, movie watching entertainments etc., an innovative business model that integrate MaaS service with the consuming services through the provision of discounted mobility service by consuming service providers after booking or finished the consuming services is proposed, and technically implemented through QR-Code produced by mobile App, which act as a connector through payment and identification. Finally, a demonstration application is carried out in the city of Guangzhou, the observed data shows that the proposed business model could empower both type of services, increase the volume of MaaS service while attracting and enlarging more consumptions.

Abstract

In this paper, a smart concrete pavement weigh-in-motion system based on the deep learning method is studied. First, a sensor network using semiconductor strain gage is designed to measure the pressure information of the vehicle which is set under the concrete pavement. T...Lead moreIn this paper, a smart concrete pavement weigh-in-motion system based on the deep learning method is studied. First, a sensor network using semiconductor strain gage is designed to measure the pressure information of the vehicle which is set under the concrete pavement. Then, a specific designed hardware is included in order to collect the data delivered to the host computer. Finally, the label system based on the traditional weighing product and the new data obtained by the proposed system are utilized. A convolutional neural network is designed to let the proposed smart system learn to weigh the vehicle on the road dynamically. The testing results validate the accuracy and efficiency of the proposed smart system.

Abstract

The integration of advanced models that focus on freight transportation, as well as that of passenger and freight agent-based models, has not been fully achieved for real-world urban transportation analysis. In this paper, we propose the framework of SimMobility Freight, ...Lead moreThe integration of advanced models that focus on freight transportation, as well as that of passenger and freight agent-based models, has not been fully achieved for real-world urban transportation analysis. In this paper, we propose the framework of SimMobility Freight, which is one of the main components of SimMobility, a multi-scale agent-based urban transportation simulation platform. We design SimMobility Freight with the aim of achieving full-scale integration of disaggregate, agent-based freight and passenger simulations, which is flexible and extendable. SimMobility Freight is capable of simulating commodity contracts, logistics planning, goods vehicle operations and parking in a fully-disaggregate manner, which allows us to predict the changes in various aspects of logistics operations and to measure their impacts. Currently, the estimated and calibrated models of SimMobility Freight are available for Singapore.

Abstract

A new magnetic marker system developed by authors are tested on public roads and the results are reported. In the experiments, magnetic markers are installed along the courses of the public roads and self-driving buses run along the maker, with the aid of other sensing de...Lead moreA new magnetic marker system developed by authors are tested on public roads and the results are reported. In the experiments, magnetic markers are installed along the courses of the public roads and self-driving buses run along the maker, with the aid of other sensing devices. Vehicle position detection accuracy is compared between marker system and RTK-GPS. Advantages of magnetic marker system is discussed. A novel RF-ID magnetic marker is proposed to provide enhance the position information including absolute positioning.

Abstract

The automatic train operation (ATO) system is an important component of the rail intelligent transport systems. Aiming at the intelligence and stability requirements of ATO control algorithm, based on fuzzy theory, gray theory and immune mechanism, a fuzzy gray immune PID...Lead moreThe automatic train operation (ATO) system is an important component of the rail intelligent transport systems. Aiming at the intelligence and stability requirements of ATO control algorithm, based on fuzzy theory, gray theory and immune mechanism, a fuzzy gray immune PID control algorithm (FGIPID) was proposed. Taking a maglev train operation control model as the controlled object, the general PID, incremental PID (INPID), fuzzy immune PID (FIPID) and FGIPID control algorithm were tested respectively by step signal and target speed curve of ATO system. The analysis results showed that the FGIPID control algorithm had an ideal step response and anti-interference ability. The output of the ATO system also had the ideal traceability, precision of parking, punctuality and comfort. The test results show the effectiveness of the proposed algorithm.

Abstract

Yamaha Motor is working on research and development of low-speed automated driving vehicles based on golf carts and control servers for centralized fleet management. The future aim is to provide mobility services for the public, particularly the elderly and people with ch...Lead moreYamaha Motor is working on research and development of low-speed automated driving vehicles based on golf carts and control servers for centralized fleet management. The future aim is to provide mobility services for the public, particularly the elderly and people with children, covering a radius of several kilometers in urban areas or resorts. This paper examines field surveys and customer value in mobility services within the Kanucha resort in Okinawa, where nearly 300 golf carts are utilized, with the aim of developing automated driving service systems for use within resorts. In addition, service model hypotheses have been drawn up based on the results of analyzing customer value, and an on-demand type service has been installed on the control servers. This paper introduces the design of the service application, the service system front end, as one specific outcome from the field surveys.

Abstract

Short-term traffic forecast is one of main themes in intelligent transportation system (ITS). For example, accurate forecast allows travellers to make appropriate travel plans and routes. To forecast precise travel time, it is a feasible way to make use of travel time on ...Lead moreShort-term traffic forecast is one of main themes in intelligent transportation system (ITS). For example, accurate forecast allows travellers to make appropriate travel plans and routes. To forecast precise travel time, it is a feasible way to make use of travel time on extensive range of road network. However, it’s still up in the air that to what extent we should use road network and how long we should observe travel times. In this paper, we show the result of an experiment to reveal fine setting for these parameters and propose a framework which forecasts travel time precisely.

Abstract

The toll system of toll roads in Japan is quite complicated, as the toll varies by the section used, the day of the week, and the time of day. It is also expected that the toll calculation method will become even more complicated if a route-based toll mechanism is introdu...Lead moreThe toll system of toll roads in Japan is quite complicated, as the toll varies by the section used, the day of the week, and the time of day. It is also expected that the toll calculation method will become even more complicated if a route-based toll mechanism is introduced based on government measures to address a “toll system to use expressways wisely.” We are introducing toll and route search systems that allow users to easily search tolls, and can be used by both road users and road operators for various applications in various situations. We will also describe the method of application and future prospects for the toll and route search systems.

Abstract

In a conventional traffic control system, traffic volume and speed are measured by vehicle detectors installed on the road, and the queue length is estimated, thereby the time of green or red traffic signal is appropriately controlled. However, since a large number of veh...Lead moreIn a conventional traffic control system, traffic volume and speed are measured by vehicle detectors installed on the road, and the queue length is estimated, thereby the time of green or red traffic signal is appropriately controlled. However, since a large number of vehicle detectors are required, high installation and maintenance costs have become a problem. On the other hand, in recent years, probe data collected from cars and smartphones has become widespread, and it has become possible to grasp traffic conditions even in places where vehicle detectors are not installed. Therefore, by utilizing the probe data as an alternative to the vehicle detector, we are aiming to realize a lower cost signal control. For this purpose, we have developed a technique to estimate the queue length required for signal control from fewer vehicle detector data and probe data using a neural network.

Abstract

It’s possible to select multiple routes to connect from the location of departure to the destination within the expressway network in Japan, and its current mechanism is designed so that the toll will remain the same if the location of departure and the destination are th...Lead moreIt’s possible to select multiple routes to connect from the location of departure to the destination within the expressway network in Japan, and its current mechanism is designed so that the toll will remain the same if the location of departure and the destination are the same. To address traffic demand management mechanisms, including toll collection based on the travelled distance and the route guidance toll collection utilizing the expressway network, which are being examined by the government and others in recent years, it is expected a grasp of the routes of the traveling vehicles will be needed. In this study, we verified and evaluated the accuracy of location identification on Japan’s expressways while using GNSS positioning technology, which is being considered as a potential method to grasp routes.

Abstract

Rapid and Accurate Incident Detection is an increasingly complex challenge for Video Automatic Incident Detection systems, especially in the context of consolidated operation of multiple high traffic volume assets. Currently available single video source technologies are ...Lead moreRapid and Accurate Incident Detection is an increasingly complex challenge for Video Automatic Incident Detection systems, especially in the context of consolidated operation of multiple high traffic volume assets. Currently available single video source technologies are unable to meet the performance requirements of these environments – creating alarm overwhelm and a significant barrier to automation of operations. To address this challenge, Transurban initiated trials based upon sensor aggregation to increase the performance of incident detection in real time: reducing false and nuisance alarms and increasing detection performance. The goal is to use multiple sources of data of varying individual performance in aggregation to provide performance at a greater level than the individual technologies, both in a desktop environment, but also rapidly prototyping the solution in production to allow greater Operational input and earlier benefit. The implementation of this solution targets enabling greater Operator efficiency as well as increasing the potential for automation, which should further increase as the performance of individual data sources is improved in parallel. This presentation will report on the outcomes of this work which is underway across FY2018/2019.

Abstract

Urban sprawl, population growth and technology driven service models are creating pressure on the transport network. The traditional approach of ‘building our way out of congestion’ is no longer effective nor sustainable. There is an ever-growing need for a more viable an...Lead moreUrban sprawl, population growth and technology driven service models are creating pressure on the transport network. The traditional approach of ‘building our way out of congestion’ is no longer effective nor sustainable. There is an ever-growing need for a more viable and different approach in managing travel demand and providing alternative choices. This paper describes how travel demand management outcomes can be achieved and the enabling mechanisms to accomplish this. The key focus of this paper will be on establishing a framework for the development of workstreams under a travel demand management programme. This includes outlining example workstreams and the associated funding allocation, delivery programme and governance structure.

Abstract

Disabled people generally want to stand on their own two feet, and achieving mobility is an important step in satisfying that desire. A steering-operation unit for disabled people with disability in their arms was developed and experimentally evaluated. The unit consists ...Lead moreDisabled people generally want to stand on their own two feet, and achieving mobility is an important step in satisfying that desire. A steering-operation unit for disabled people with disability in their arms was developed and experimentally evaluated. The unit consists of a set of load-cell sensors, one for turning right and one for turning left. The driver steps the right or left load cell to turn the car right or left. The magnitude of the driver’s stepping force is converted to a voltage and input to the power-steering motor. The angular velocity of the steering wheel corresponds to that voltage. As a result of this configuration, the driver can drive a car just by moving their foot and intuitively selects the load-cell they must apply by foot to turn the car. Experimental results using a standard car fitted with the developed steering operation unit show that disabled people can drive the car with their foot in a manner close to that achieved with a steering wheel.

Abstract

According to the requirement of congestion charging based on ETC system, this paper designs a MLFF RSU controller based on national standard DSRC protocol, and cooperates with the ETC RSU equipment based on space array signal processing. The main technical measures includ...Lead moreAccording to the requirement of congestion charging based on ETC system, this paper designs a MLFF RSU controller based on national standard DSRC protocol, and cooperates with the ETC RSU equipment based on space array signal processing. The main technical measures include: supporting special PCI interface encryption card to eliminate data processing speed bottleneck; using high-speed optical fiber communication to reduce communication time between RSU and the controller; using high-performance independent ARM processor to complete parallel transaction process. After a series of development, through the actual test, the MLFF RSU controller designed in this paper can reduce the data transaction time by 40% compared with the traditional ETC RSU. In conclusion, it shows that the MLFF RSU controller designed in this paper meets the platform optimization requirements of congestion toll application and achieves the expected design purpose.

Abstract

This paper focused on the potential support evaluation of Bus Rapid Transit (BRT) for commuting activities by using mobile phone signaling data. Algorithms that could identify the living places and workplaces of mobile phone users were proposed. A case study in Urumqi, Ch...Lead moreThis paper focused on the potential support evaluation of Bus Rapid Transit (BRT) for commuting activities by using mobile phone signaling data. Algorithms that could identify the living places and workplaces of mobile phone users were proposed. A case study in Urumqi, China was given to apply the algorithms and validate the framework. Four performance indicators were designed and calculated. The results indicated that both half of living places and workplaces in Urumqi were covered by the service of BRT systems with a buffer of 1000 meters. About 50% of the working population who lived within the service buffer also worked at this area. There were over 65% of the residents living within the service buffer whose working places were also falling in the service area of the BRT system.

Abstract

Inattentive and aggressive driving behaviors are crucial factors to deteriorate the safety performance of traffic stream by crash occurrence and shock wave generation. For these reasons, various countermeasures are under development to modify driving behavior toward crash...Lead moreInattentive and aggressive driving behaviors are crucial factors to deteriorate the safety performance of traffic stream by crash occurrence and shock wave generation. For these reasons, various countermeasures are under development to modify driving behavior toward crash prevention and severity reduction. This study attempts to identify the impact of driving behavior modification, which can be realized by advanced driving assistance systems (ADAS), on the traffic stream in terms of safety performance. Microscopic traffic simulation experiments using VISSIM were conducted to analyze driving behavior parameters associated with the longitudinal and lateral vehicle maneuvering. The VISSIM COM interface was further used to implement a genetic algorithm (GA) optimizer to identify the maximum achievable safety benefits by modifying driving behavior. In addition, various market penetration rates (MPR) of ADAS-equipped vehicles were applied into the analysis framework. Results showed that optimal parameter combinations to minimize the acceleration noise varied by different MPR levels. It was, for example, observed that approximately 12% reduction of acceleration noise was obtainable at 60% MPR level. The outcome of this study would be useful in establishing safety policies and supporting technological development toward the safety enhancement by behavior modification.

Abstract

For any country to successfully shift more commute trips to public transport, besides network coverage and scheduling regularity, local authorities are cognizant of the importance of service reliability, travel comfort and journey times. Quality and performance of PT mode...Lead moreFor any country to successfully shift more commute trips to public transport, besides network coverage and scheduling regularity, local authorities are cognizant of the importance of service reliability, travel comfort and journey times. Quality and performance of PT modes need to attain a state where the user’s perceived utilities with PT exceeds that associated with more private modes. In Singapore, our land scarcity and dense network accentuates the importance of PT efficiency, which translates to smoother bus journeys and shorter PT travel times. Comprehensive packages and regulatory monitoring have been in place to ensure regularity and dependency on our bus journeys, as well as parallel efforts to improve bus speeds. This study seeks to delve deep into the factors affecting bus speeds, using advanced statistical tool to understand the influence weightage of the key root causes of low bus speeds. It does this by harnessing the rich raw data sets collected by the Land Transport Authority’s myriad of Intelligent Transport Systems, covering performance measures from both general traffic and bus transit modes. These data are integrated for the first time in an innovative way, into a cohesive data-set through data-layering and common identifiers, to answer Singapore’s unique PT challenges.

Abstract

In this paper, the vehicle information sharing (VIS) performance at blind intersection is analysed from the viewpoint of the V2V on-board unit (OBU) penetration rate. a microscopic traffic simulator capable of simulating the MAC layer was constructed. we analyze the influ...Lead moreIn this paper, the vehicle information sharing (VIS) performance at blind intersection is analysed from the viewpoint of the V2V on-board unit (OBU) penetration rate. a microscopic traffic simulator capable of simulating the MAC layer was constructed. we analyze the influence of MAC of CSMA on the VIS performance by the simulator.

Abstract

Traffic surveyors have traditionally used fixed cameras as the primary data collection method to study the operational efficiency of signalised intersections. With recent advancements in Unmanned Aerial Vehicles (UAVs) technology, this study explores the viability of UAV ...Lead moreTraffic surveyors have traditionally used fixed cameras as the primary data collection method to study the operational efficiency of signalised intersections. With recent advancements in Unmanned Aerial Vehicles (UAVs) technology, this study explores the viability of UAV for traffic surveying applications in New Zealand. The paper reports on a traffic survey conducted using UAVs to collect data from a signalized intersection in Auckland. The aerial video data is processed to estimate the observed queue length and control delays and compared against those calculated using two analytical models namely: shockwave model and deterministic queuing model and SIDRA Intersection 7 Software (modified ARR 123 & HCM settings). The results from this study suggest that UAV traffic surveying is a viable method which can provide very accurate data in a cost-effective manner. However, with the current UAV regulations in New Zealand, the number of available study locations is heavily limited. As this is the first study in New Zealand which investigates UAV technology for transportation engineering applications, the results and proposed methodology may act as a benchmark and framework for future researchers and practitioners alike.

Abstract

In this paper, we propose a data-driven modelling approach with a novel hierarchical D-CLSTM-t deep learning model for short-term traffic speed prediction, a framework combined with convolutional neural network (CNN) and long short-term memory (LSTM) models. A deep CNN mo...Lead moreIn this paper, we propose a data-driven modelling approach with a novel hierarchical D-CLSTM-t deep learning model for short-term traffic speed prediction, a framework combined with convolutional neural network (CNN) and long short-term memory (LSTM) models. A deep CNN model is employed to learn the spatio-temporal traffic patterns of the input graphs, which are subsequently fed into a deep LSTM model for sequence learning. To capture traffic seasonal variations, time of the day and day of the week indicators are fused with trained features from the CNN-LSTM framework. The model is trained end-to-end to predict travel speed in 15 to 90 minutes in the future. We compare the model performance against other baseline models including CNN, LGBM, LSTM, and traditional speed-flow curves. Experiment results show that the DCLSTM-t outperforms other models considerably. Model tests show that speed upstream also responds sensibly to a sudden accident occurring downstream. Our D-CLSTM-t model framework is also highly scalable for future extension such as for network-wide traffic prediction, which can also be improved by including additional features such as weather, long term seasonality and accident information.

Abstract

Traffic prediction is a critical component in intelligent traffic control systems to project future vehicle evolution in order to make optimal decisions for all competing demand in real-time. In this paper, we study the use of cell transmission model for traffic predictio...Lead moreTraffic prediction is a critical component in intelligent traffic control systems to project future vehicle evolution in order to make optimal decisions for all competing demand in real-time. In this paper, we study the use of cell transmission model for traffic prediction in signalized urban network. Existing models treat traffic flow on a link as a single commodity until the flow reaches diverge point or turn bay. Turning ratio is applied at the diverge point to compute the flow going to individual movement. Lane blockage and spill-back happens due to insufficient green time or imperfect road structure such as short turn bay or road incidents. Most likely the effect of lane blockage and spill-back will not have equal influence on every movement. To address this, we extend cell transmission model to keep track of vehicle route intent in every cell and perform vehicle progression at the movement level. To evaluate feasibility and accuracy of the model for use in real-time, this study has developed a complete modeling framework comprising connected vehicle traffic simulation platform, real-time traffic tracking and prediction platforms and has conducted extensive simulation experiments with a segment of Corporation Road and Boon Lay Way in Jurong West, Singapore.

Abstract

Despite the increasing number of studies on various factors that influence public transportation modes, there is little knowledge of their relative contribution to explain variance in bike sharing ridership in a city. In this study, we present the concept of Bike Sharing ...Lead moreDespite the increasing number of studies on various factors that influence public transportation modes, there is little knowledge of their relative contribution to explain variance in bike sharing ridership in a city. In this study, we present the concept of Bike Sharing kilometres Travelled (BSKT) and propose a method for its estimation using more than 5 Million trip trajectories of GPS collected from the bike sharing system of the city of Seoul. BSKT Ridership models based on multiple linear regression, Bayesian regression and a conditional autoregressive model were developed and compared to understand the combined contributions of several categories of factors on BSKT variability in the city of Seoul, i.e., bike sharing specific, Road network and topology, socio-economic, land use, and built environment variables. The variables were aggregated at the traffic analysis zone (TAZ) level. Results suggest a strong correlation mainly between BSKT and bike sharing stations and bicycle network variables. In addition, it can be shown that bike sharing stations, dedicated bicycle roads, dense activity land use, and better transit station coverage contribute to the increase of BSKT while poor roads, hills, high income areas and safety for both road and land use types have negative impacts on BSKT.

Abstract

With the rapid development of urbanization and traffic mobility, problems about energy consumption and air pollution caused by road traffic are getting more and more serious. To achieve the purpose of energy saving and emission reduction of urban transport, a relatively s...Lead moreWith the rapid development of urbanization and traffic mobility, problems about energy consumption and air pollution caused by road traffic are getting more and more serious. To achieve the purpose of energy saving and emission reduction of urban transport, a relatively stable driving cycle can be proposed in addition to vehicle use restriction and fuel efficiency improvement. This paper attempted to put forward a spatial driving cycle construction method of left-turn vehicles at signalized intersections to achieve the purpose of energy saving and emission reduction of urban transport by obtaining a relatively stable driving cycle. First, the paper analysed the spatial characteristics of driving cycle data of the spatial driving characteristics. Then the state transition matrix is calculated by Markov method and the spatial driving cycle is constructed by driving state selection and speed splicing method, including ontology construction and general construction. Finally, the fitting degree of the results is evaluated based on the determination coefficient of the average speed and the error of the joint distribution of velocity-acceleration, and the result shows that the constructed driving cycle is similar with the measured driving cycle in spatial series.

Abstract

This paper considers the vehicle routing problem of a demand-responsive transport service that can cater both passengers and freight. The vehicles in the service are shared among these two types of customers. This paper offers an optimisation formulation of the problem wi...Lead moreThis paper considers the vehicle routing problem of a demand-responsive transport service that can cater both passengers and freight. The vehicles in the service are shared among these two types of customers. This paper offers an optimisation formulation of the problem with a comprehensive objective function, which considers the trade-off between service level and operating costs. In addition, the formulation does not allow demand rejection by using soft time-window constraints and attempts to minimise the fleet size at the same time. Furthermore, this paper offers the first application of an anticipatory algorithm for routing optimisation of such services. The benefits of using a particular anticipatory algorithm, the Multiple Scenario Approach (MSA), are studied and the solution is compared to the solutions produced by other techniques, namely greedy insertion and an offline deterministic benchmark case (where all demands are assumed to be known a priori). These algorithms are implemented in an event-based vehicle routing simulation experiment by using a demand scenario obtained from the Victorian Integrated Survey of Travel and Activity (VISTA) data. The experiment results highlight the benefits of using the comprehensive objective function along with the MSA to improve the service level while still maintaining a reasonable cost.

Abstract

Each traveller recognizes routes and travel times based on their experience of travel, and travelers make different choices about routes based on their perceived travel time. The purpose of this study is to classify travelers using risk appetite to analyse path selection ...Lead moreEach traveller recognizes routes and travel times based on their experience of travel, and travelers make different choices about routes based on their perceived travel time. The purpose of this study is to classify travelers using risk appetite to analyse path selection behaviour. The risk preference is a behavioural characteristic of the on-time arrival according to the probable travel time and is used to classify the travelers into four categories. Daegu’s DSRC (Dedicated Short Range Communication) data was used to analyse route selection behaviour by 7,018 observers and each traveller was observed more than 3 times at a specific departure point. We construct a selection set generation model to select four of the 16 significant routes. Probabilistic Selection The set generation model is designed and developed using the probabilistic nature of path travel time. The path selection model is set for each group and the path selection behaviour is analysed using the multinomial logit model (MNL). We use service level and network attribute variables for analysis. Statistical analysis shows that there are different path selection behaviours among the four groups. Other behaviours in path selection were analysed by other factor sizes to analyse the heterogeneity of the tourist population. This study contributes to the development of various strategies for obtaining information by taking into account the path selection behaviour of each group.

Abstract

Abstract: Urban traffic management requires dynamic traffic demand data, like dynamic origin-destination (O-D) flows, to be adapted to the rapid and short-term variation of dynamic traffic flows. In arterials, travel time considering signal control is an important element...Lead moreAbstract: Urban traffic management requires dynamic traffic demand data, like dynamic origin-destination (O-D) flows, to be adapted to the rapid and short-term variation of dynamic traffic flows. In arterials, travel time considering signal control is an important element affecting the precision of dynamic O-D estimation. This paper proposed a revised dynamic O-D estimation method for arterials based on Kalman filtering (KF) which included three steps “pre-estimation, simulation and revisement”. In simulation process, varied signal control timing was considered into travel time calculation. Moreover, revised dynamic travel time with respect to signal control was used to revise KF measure equation. As a result, precision of dynamic O-D estimation for arterials was ensured. Case Study with abundant error analyses between real data and estimated data of an arterial in Beijing proved the effectiveness of this method in acquiring dynamic O-D flows in arterials and it will be helpful for urban traffic management in bright future.

Abstract

Traffic incidents usually lead to very severe congestion in urban networks. In light of this, an effective traffic incident detection method is one of the key solutions in the intelligent transportation system as the guide of traffic control to manage traffic congestion t...Lead moreTraffic incidents usually lead to very severe congestion in urban networks. In light of this, an effective traffic incident detection method is one of the key solutions in the intelligent transportation system as the guide of traffic control to manage traffic congestion timely and specifically. This paper presents a Convolutional Neural Network (CNN) based traffic incident detection method to detect the incidents in urban networks by three types of traffic flow data (traffic volume, speed and acceleration of vehicles). Regarding the traffic flow data during a time period as an image with RGB channels, this method employs the idea of CNN for dealing with image classification. For testing the performance, this paper develops an incident generation simulation model on a microscopic simulation platform VISSIM to generate adequate 3 classes traffic flow data, where the traffic demand of each simulation is estimated on the basis of true traffic flow data collected by the SCATS system in Singapore. Three experiments show the proposed method can effectively and efficiently detect even minor incidents in urban networks and the acceleration data has the biggest impact on the detection accuracy. Comparing with LSTM and BP methods, the CNN-based method also demonstrates its high performance.

Abstract

Transit has played an important role to minimize the traffic congestion and reduce the travel costs in urban area. In addition to this role, it is necessary to evaluate the efficiency of transit operation from the transit demand considering route overlapping. This study a...Lead moreTransit has played an important role to minimize the traffic congestion and reduce the travel costs in urban area. In addition to this role, it is necessary to evaluate the efficiency of transit operation from the transit demand considering route overlapping. This study aims to analyze bus trips of overlapping O-D pairs to enhance the efficiency of bus operations. Three CBDs were selected as the area and peak hour were selected as the time to consider passenger’s point of view. We calculate knee point of the travel demand through the Kneedle algorithm and extract demands over the knee point. We find the overlapping O-D pairs though the travel demand. The demand from O-D pairs are statistically analyzed and visualized on a map. It concludes that demand-based overlap index and segment-based overlap index are not related to each other by calculating correlation coefficient. The proposed procedure is applied with combining the average bus dispatch time. This proposed approach enables operators to evaluate overlap bus route considering demand. It has significance for the passenger to increase the efficiency in evaluating bus systems and planning bus route using this method.

Abstract

In a urban logistic system, the prediction of next destination and estimated time of arrival (ETA) for delivery fleets is of paramount importance for efficient resource planning and providing satisfactory client experience. The level of prediction is limited by the inform...Lead moreIn a urban logistic system, the prediction of next destination and estimated time of arrival (ETA) for delivery fleets is of paramount importance for efficient resource planning and providing satisfactory client experience. The level of prediction is limited by the information accessible to individual logistic companies or warehouse operators, and further complicated by the complex urban road system. Nonethless, data collection and synergy under smart city initiatives worldwide provides exhilarating new opportunities to breakthrough these limitations. In this paper, we propose a data driven framework for next destination and ETA prediction of delivery fleets based on data collected in Singapore, which is subjected to mature urban infrastructures and featured by its smart nation initiative. The proposed framework consists of two main components: a next destination and ETA predictor based on delivery fleet’s historical Global Positioning System (GPS) trajectory data using nonlinear autoregressive neural network (NARNN), and a real-time incident detector based on urban traffic data using a random forest model, which results can be used for further delay estimation and route optimization.

Abstract

Logistics agencies have faced efficiency challenges in fleet management because of rapid growth in an economy. They need an efficient method to support their decision making. In this study, an incident model is proposed to analyze the cause and consequence of an incident ...Lead moreLogistics agencies have faced efficiency challenges in fleet management because of rapid growth in an economy. They need an efficient method to support their decision making. In this study, an incident model is proposed to analyze the cause and consequence of an incident that occurred during January through October 2018. A Bayesian network was used, and a Bayesian neural network was proposed to detect incidents from obtained knowledge and make a prediction on loss of profits. The focus was on the shortage of vehicle incident information. The study showed that the incident could not be analyzed logically, because the incident was uncertainty and it did not contain any regular pattern. However, the inefficient fleet management caused a shortage of vehicle incident for demand handling (0.74 in probability). Furthermore, each time incidents happened, the loss in profit was at least 11,774.57 THB on average per day and 353,237 THB per month. This study contributed an incident model in which one can represent knowledge for recognizing the causes and consequences of incidents and support decision making in fleet vehicle management at a fundamental level.

Abstract

In this paper, we build a Long Short Term Memory (LSTM) model to predict the demand for taxi service in Kaohsiung City, Taiwan, where the prediction is station-based and given for the next 30 minutes. The training data of the model mainly consist of demand count, weather ...Lead moreIn this paper, we build a Long Short Term Memory (LSTM) model to predict the demand for taxi service in Kaohsiung City, Taiwan, where the prediction is station-based and given for the next 30 minutes. The training data of the model mainly consist of demand count, weather conditions, and local pedestrians. Since a taxi station takes in charge of a dispatch-area with an irregular shape, the data are allocated to the belonging stations before training. In addition, when transforming data into time-series data, the time-step is defined as 30-min to unify the data of various time-formats from different resources. In experiments, we build 5 models with different features and also we select 8 dispatch-areas as the targets to evaluate the model. The results show that the best sMAPE of the LSTM models is 0.195 and the average error of overall 8 dispatch-areas is 0.28.

Abstract

This paper presents a novel method to estimate the queue at signalised intersections in a connected vehicle (CV) environment. It is a parsimonious estimation method with no need of any conventional information such as arrival flow and parameters of traffic signals. To thi...Lead moreThis paper presents a novel method to estimate the queue at signalised intersections in a connected vehicle (CV) environment. It is a parsimonious estimation method with no need of any conventional information such as arrival flow and parameters of traffic signals. To this end, we propose the idea of “a leader CV” to minimise the communication load and reduce the risk of communication failure which is a serious concern in CV ecosystems. Furthermore, a linear regression is used to weigh the importance of input variables in a neural network based queue prediction model. Vissim traffic simulator is deployed to train and then evaluate the effectiveness and robustness of the model under different travel demand conditions, varying number of CVs (i.e., CVs’ market penetration rate) and various traffic signal scenarios. As expected, when the market penetration rate increases, the accuracy of the model enhances consequently. In a congested traffic condition, the proposed model is more accurate compared to the under-saturated condition with the same market penetration rate. Although the proposed method does not depend on arrival information of vehicles and traffic signal parameters, the accuracy of the queue estimation results is still comparable with methods that highly depend on such information

Abstract

Automated vehicles are expected to be released in the automobile market by 2020 and roads will be inevitably shared by manually driven vehicles and automated vehicles. In this situation the driving behavior of the surrounding vehicles must be identified to facilitate safe...Lead moreAutomated vehicles are expected to be released in the automobile market by 2020 and roads will be inevitably shared by manually driven vehicles and automated vehicles. In this situation the driving behavior of the surrounding vehicles must be identified to facilitate safe driving of automated vehicles, especially when they are faced with a mandatory lane-changing. For this reason, a methodology that can evaluate the driving riskiness of vehicles placed surrounding lanes from the relative speed of car-following was developed. The relative speed in car-following is shown in waveforms can be decomposed into harmonics, which are subordinate waves, by performing a spectral analysis using Fourier transform. The components showing the riskiness of following vehicles among these harmonics were derived through a correlation analysis with adjusted time-to-collision. Consequently, the power spectrum density ratio with a frequency band of over 0.017Hz was analyzed to have a high correlation with riskiness, and was defined as the risk propensity index.