Hierarchical travel demand estimation using multiple data sources: A forward and backward propagation algorithmic framework on a layered computational graph

Aiming to develop a theoretically consistent framework to estimate travel demand using multiple data sources, this paper first proposes a multi-layered Hierarchical Flow Network (HFN) representation to structurally model different levels of travel demand variables including trip generation, origin/destination matrices, path/link flows, and individual behavior parameters. Different data channels from household travel surveys, smartphone type devices, global position systems, and sensors can be mapped to different layers of the proposed network structure. We introduce Big data-driven Transportation Computational Graph (BTCG), alternatively Beijing Transportation Computational Graph, as the underlying mathematical modeling tool to perform automatic differentiation on layers of composition functions. A feedforward passing on the HFN sequentially implements 3 steps of the traditional 4-step process: trip generation, spatial distribution estimation, and path flow-based traffic assignment, respectively. BTCG can aggregate different layers of partial first-order gradients and use the back-propagation of “loss errors” to update estimated demand variables. A comparative analysis indicates that the proposed methods can effectively integrate different data sources and offer a consistent representation of demand. The proposed methodology is also evaluated under a demonstration network in a Beijing subnetwork.     

Analytical model for information flow propagation wave under an information relay control strategy in a congested vehicle-to-vehicle communication environment

Vehicular traffic congestion in a vehicle-to-vehicle (V2V) communication environment can lead to congestion effects for information flow propagation. Such congestion effects can impact whether a specific information packet of interest can reach a desired location, and if so, in a timely manner to influence the traffic system performance. Motivated by the usefulness and timeliness of information propagation, this paper aims to characterize the information flow propagation wave (IFPW) for an information packet in a congested V2V communication environment under an information relay control strategy. This strategy seeks to exclude information that is dated in the communication buffer under a first-in, first-out queue discipline, from being relayed if the information flow regime is congested. It trades off the need to enable the dissemination of every information packet as far as possible, against the congestion effects that accrue because of the presence of multiple information packets. A macroscopic two-layer model is proposed to characterize the IFPW. The upper layer is formulated as integro-differential equations to characterize the information dissemination in space and time under this control strategy. The lower layer adopts the Lighthill-Whitham-Richards model to capture the traffic flow dynamics. Based on the upper layer model, a necessary condition is derived which quantifies the expected time length that needs to be reserved for broadcasting the information packet of interest so as to ensure the formation of an IFPW under a given density of V2V-equipped vehicles. When the necessary condition is satisfied under homogeneous conditions, it is shown that the information packet can be propagated at an asymptotic speed whose value can be derived analytically. Besides, under the proposed control strategy, only a proportion of vehicles (labeled asymptotic density of informed vehicles) can receive the specific information packet, which can be estimated by solving a nonlinear equation. The asymptotic IFPW speed, the asymptotic density of informed vehicles, and the necessary condition for the IFPW, help in evaluating the timeliness of information propagation and the influence of traffic dynamics on information propagation. In addition, the proposed model can be used to numerically estimate the IFPW speed for heterogeneous conditions, which can aid in the design of traffic management strategies built upon the timely propagation of information through V2V communication.     

Stochastic dynamic switching in fixed and flexible transit services as market entry-exit real options

The first analytical stochastic and dynamic model for optimizing transit service switching is proposed for “smart transit” applications and for operating shared autonomous transit fleets. The model assumes a region that requires many-to-one last mile transit service either with fixed-route buses or flexible-route, on-demand buses. The demand density evolves continuously over time as an Ornstein-Uhlenbeck process. The optimal policy is determined by solving the switching problem as a market entry and exit real options model. Analysis using the model on a benchmark computational example illustrates the presence of a hysteresis effect, an indifference band that is sensitive to transportation system state and demand parameters, as well as the presence of switching thresholds that exhibit asymmetric sensitivities to transportation system conditions. The proposed policy is computationally compared in a 24-hour simulation to a “perfect information” set of decisions and a myopic policy that has been dominant in the flexible transit literature, with results that suggest the proposed policy can reduce by up to 72% of the excess cost in the myopic policy. Computational experiments of the “modular vehicle” policy demonstrate the existence of an option premium for having flexibility to switch between two vehicle sizes.     

Estimation of flow and density using probe vehicles with spacing measurement equipment

Probe vehicles provide some of the most useful data for road traffic monitoring because they can acquire wide-ranging and spatiotemporally detailed information at a relatively low cost compared with traditional fixed-point observation. However, current GPS-equipped probe vehicles cannot directly provide us volume-related variables such as flow and density. In this paper, we propose a new probe vehicle-based estimation method for obtaining volume-related variables by assuming that a probe vehicle can measure the spacing to its leading one. This assumption can be realized by utilizing key technologies in advanced driver assistance systems that are expected to spread in the near future. We developed a method of estimating the flow, density, and speed from the probe vehicle data without exogenous assumptions on traffic flow characteristics, such as a fundamental diagram. In order to quantify the characteristics of the method, we performed a field experiment at a real-world urban expressway by employing prototypes of the probe vehicles with spacing measurement equipment. The result showed that the proposed method could accurately estimate the 5 min and hourly traffic volumes with probe vehicle penetration rate of 3.5% and 0.2%, respectively.     

Fueling for contingencies: The hidden cost of unpredictability in the air transportation system

The aviation community is increasing its attention on the concept of predictability when conducting aviation service quality assessments. Reduced fuel consumption and the related cost is one of the various benefits that could be achieved through improved flight predictability. A lack of predictability may cause airline dispatchers to load more fuel onto aircraft before they depart; the flights would then in turn consume extra fuel just to carry excess fuel loaded. In this study, we employ a large dataset with flight-level fuel loading and consumption information from a major US airline. With these data, we estimate the relationship between the amount of loaded fuel and flight predictability performance using a statistical model. The impact of loaded fuel is translated into fuel consumption and, ultimately, fuel cost and environmental impact for US domestic operations. We find that a one-minute increase in the standard deviation of airborne time leads to a 0.88 min increase in loaded contingency fuel and 1.66 min in loaded contingency and alternate fuel. If there were no unpredictability in the aviation system, captured in our model by eliminating standard deviation in flight time, the reduction in the loaded fuel would between 6.12 and 11.28 min per flight. Given a range of fuel prices, this ultimately would translate into cost savings for US domestic airlines on the order of $120–$452 million per year.     

A study of feasibility and potential benefits of organised car sharing in Ireland

In this study, the market potential of car sharing has been evaluated using multiple alternative scenarios which examine the geographic, financial and environmental factors influencing car sharing adoption. The scenarios are applied to the available and collected travel information of the Irish population to estimate the potential impact of introducing car sharing in Ireland. The analysis identified that car owners who travel predominantly on alternative modes, could make significant cost and CO₂ savings through car sharing. A reduction of yearly CO₂ emissions of 86 kt is readily achievable through car sharing, with reductions up to 895 kt possible with appropriate policy and financial support. These figures are comparable to other measures proposed under the Irish National Climate Change Strategy.     

Network-oriented household activity pattern problem for system optimization

The recently emerging trend of self-driving vehicles and information sharing technologies, made available by private technology vendors, starts creating a revolutionary paradigm shift in the coming years for traveler mobility applications. By considering a deterministic traveler decision making framework at the household level in congested transportation networks, this paper aims to address the challenges of how to optimally schedule individuals’ daily travel patterns under the complex activity constraints and interactions. We reformulate two special cases of household activity pattern problem (HAPP) through a high-dimensional network construct, and offer a systematic comparison with the classical mathematical programming models proposed by Recker (1995). Furthermore, we consider the tight road capacity constraint as another special case of HAPP to model complex interactions between multiple household activity scheduling decisions, and this attempt offers another household-based framework for linking activity-based model (ABM) and dynamic traffic assignment (DTA) tools. Through embedding temporal and spatial relations among household members, vehicles and mandatory/optional activities in an integrated space-time-state network, we develop two 0–1 integer linear programming models that can seamlessly incorporate constraints for a number of key decisions related to vehicle selection, activity performing and ridesharing patterns under congested networks. The well-structured network models can be directly solved by standard optimization solvers, and further converted to a set of time-dependent state-dependent least cost path-finding problems through Lagrangian relaxation, which permit the use of computationally efficient algorithms on large-scale high-fidelity transportation networks.     

Simulation based population synthesis

Microsimulation of urban systems evolution requires synthetic population as a key input. Currently, the focus is on treating synthesis as a fitting problem and thus various techniques have been developed, including Iterative Proportional Fitting (IPF) and Combinatorial Optimization based techniques. The key shortcomings of these procedures include: (a) fitting of one contingency table, while there may be other solutions matching the available data (b) due to cloning rather than true synthesis of the population, losing the heterogeneity that may not have been captured in the microdata (c) over reliance on the accuracy of the data to determine the cloning weights (d) poor scalability with respect to the increase in number of attributes of the synthesized agents. In order to overcome these shortcomings, we propose a Markov Chain Monte Carlo (MCMC) simulation based approach. Partial views of the joint distribution of agent’s attributes that are available from various data sources can be used to simulate draws from the original distribution. The real population from Swiss census is used to compare the performance of simulation based synthesis with the standard IPF. The standard root mean square error statistics indicated that even the worst case simulation based synthesis (SRMSE = 0.35) outperformed the best case IPF synthesis (SRMSE = 0.64). We also used this methodology to generate the synthetic population for Brussels, Belgium where the data availability was highly limited.     

Improved driver responses at intersections with red signal countdown timers

Traffic Signal Countdown Timers (TSCTs) are innovative, practical and cost effective technologies with the potential to improve efficiency at signalized intersections. The purpose of these devices is to assist motorists in decision-making at signalized intersections with real-time signal duration information. This study focused specifically on driver responses in the presence of a Red Signal Countdown Timer (RSCT). A Linear Mixed Effect (LME) model was developed to predict the effect of RSCT on the headway of the first vehicle waiting on a red signal. The model predicted 0.72 s reduction in the headway of the first queued vehicle resulting from the presence of RSCT, while the observed difference in mean headway was 0.82 s. This result is suggestive of a reduction in start-up lost time at signalized intersections, i.e., an improvement in signalized intersection efficiency when an RSCT is present.     

Forecasting short-term subway passenger flow under special events scenarios using multiscale radial basis function networks

Reliable and accurate short-term subway passenger flow prediction is important for passengers, transit operators, and public agencies. Traditional studies focus on regular demand forecasting and have inherent disadvantages in predicting passenger flows under special events scenarios. These special events may have a disruptive impact on public transportation systems, and should thus be given more attention for proactive management and timely information dissemination. This study proposes a novel multiscale radial basis function (MSRBF) network for forecasting the irregular fluctuation of subway passenger flows. This model is simplified using a matching pursuit orthogonal least squares algorithm through the selection of significant model terms to produce a parsimonious MSRBF model. Combined with transit smart card data, this approach not only exhibits superior predictive performance over prevailing computational intelligence methods for non-regular demand forecasting at least 30 min prior, but also leverages network knowledge to enhance prediction capability and pinpoint vulnerable subway stations for crowd control measures. Three empirical studies with special events in Beijing demonstrate that the proposed algorithm can effectively predict the emergence of passenger flow bursts.     

Characterization and prediction of air traffic delays

This paper presents a new class of models for predicting air traffic delays. The proposed models consider both temporal and spatial (that is, network) delay states as explanatory variables, and use Random Forest algorithms to predict departure delays 2–24 h in the future. In addition to local delay variables that describe the arrival or departure delay states of the most influential airports and links (origin–destination pairs) in the network, new network delay variables that characterize the global delay state of the entire National Airspace System at the time of prediction are proposed. The paper analyzes the performance of the proposed prediction models in both classifying delays as above or below a certain threshold, as well as predicting delay values. The models are trained and validated on operational data from 2007 and 2008, and are evaluated using the 100 most-delayed links in the system. The results show that for a 2-h forecast horizon, the average test error over these 100 links is 19% when classifying delays as above or below 60 min. Similarly, the average over these 100 links of the median test error is found to be 21 min when predicting departure delays for a 2-h forecast horizon. The effects of changes in the classification threshold and forecast horizon on prediction performance are studied.     

Using nonparametrics to specify a model to measure the value of travel time

Using a range of nonparametric methods, the paper examines the specification of a model to evaluate the willingness-to-pay (WTP) for travel time changes from binomial choice data from a simple time–cost trading experiment. The analysis favours a model with random WTP as the only source of randomness over a model with fixed WTP which is linear in time and cost and has an additive random error term. Results further indicate that the distribution of log WTP can be described as a sum of a linear index fixing the location of the log WTP distribution and an independent random variable representing unobserved heterogeneity. This formulation is useful for parametric modelling. The index indicates that the WTP varies systematically with income and other individual characteristics. The WTP varies also with the time difference presented in the experiment which is in contradiction of standard utility theory.     

Reconstructing maximum likelihood trajectory of probe vehicles between sparse updates

Data from connected probe vehicles can be critical in estimating road traffic conditions. Unfortunately, current available data is usually sparse due to the low reporting frequency and the low penetration rate of probe vehicles. To help fill the gaps in data, this paper presents an approach for estimating the maximum likelihood trajectory (MLT) of a probe vehicle in between two data updates on arterial roads. A public data feed from transit buses in the city of San Francisco is used as an example data source. Low frequency updates (at every 200 m or 90 s) leaves much to be inferred. We first estimate travel time statistics along the road and queue patterns at intersections from historical probe data. The path is divided into short segments, and an Expectation Maximization (EM) algorithm is proposed for allocating travel time statistics to each segment. Then the trajectory with the maximum likelihood is generated based on segment travel time statistics. The results are compared with high frequency ground truth data in multiple scenarios, which demonstrate the effectiveness of the proposed approach, in estimating both the trajectory while moving and the stop positions and durations at intersections.     

A big data approach for clustering and calibration of link fundamental diagrams for large-scale network simulation applications

Existing methods for calibrating link fundamental diagrams (FDs) often focus on a limited number of links and use grouping strategies that are largely dependent on roadway physical attributes alone. In this study, we propose a big data-driven two-stage clustering framework to calibrate link FDs for freeway networks. The first stage captures, under normal traffic state, the variations of link FDs over multiple days based on which links are clustered in the second stage. Two methods, i.e. the standard k-means algorithm combined with hierarchical clustering and a modified hierarchical clustering based on the Fréchet distance, are applied in the first stage to obtain the FD parameter matrix for each link. The calibrated matrices are input into the second stage where the modified hierarchical clustering is re-employed as a static approach resulting in multiple clusters of links. To further consider the variations of link FDs, the static approach is extended by modifying the similarity measure through the principle component analysis (PCA). The resulting multivariate time-series clustering models the distributions of the FD parameters as a dynamic approach. The proposed framework is applied on the Melbourne freeway network using one-year worth of loop detector data. Results have shown that (a) similar roadway physical attributes do not necessarily result in similar link FDs, (b) the connectivity-based approach performs better in clustering link FDs as compared with the centroid-based approach, and (c) the proposed framework helps achieving a better understanding of the spatial distribution of links with similar FDs and the associated variations and distributions of the FD parameters.     

Wireless charging in California: Range,recharge, and vehicle electrification

This research evaluated the potential for wireless dynamic charging (charging while moving) to address range and recharge issues of modern electric vehicles by considering travel to regional destinations in California. A 200-mile electric vehicle with a real range of 160 miles plus 40 miles reserve was assumed to be used by consumers in concert with static and dynamic charging as a strict substitute for gasoline vehicle travel. Different combinations of wireless charging power (20–120 kW) and vehicle range (100–300 miles) were evaluated. One of the results highlighted in the research indicated that travel between popular destinations could be accomplished with a 200-mile EV and a 40 kW dynamic wireless charging system at a cost of about $2.5 billion. System cost for a 200-mile EV could be reduced to less than $1 billion if wireless vehicle charging power levels were increased to 100 kW or greater. For vehicles consuming 138 kWh of dynamic energy per year on a 40 kW dynamic system, the capital cost of $2.5 billion plus yearly energy costs could be recouped over a 20-year period at an average cost to each vehicle owner of $512 per year at a volume of 300,000 vehicles or $168 per year at a volume of 1,000,000 vehicles. Cost comparisons of dynamic charging, increased battery capacity, and gasoline refueling were presented. Dynamic charging, coupled with strategic wayside static charging, was shown to be more cost effective to the consumer over a 10-year period than gasoline refueling at $2.50 or $4.00 per gallon. Notably, even at very low battery prices of $100 per kWh, the research showed that dynamic charging can be a more cost effective approach to extending range than increasing battery capacity.     

Socially optimal electric driving range of plug-in hybrid electric vehicles

This study determines the optimal electric driving range of plug-in hybrid electric vehicles (PHEVs) that minimizes the daily cost borne by the society when using this technology. An optimization framework is developed and applied to datasets representing the US market. Results indicate that the optimal range is 16 miles with an average social cost of $3.19 per day when exclusively charging at home, compared to $3.27 per day of driving a conventional vehicle. The optimal range is found to be sensitive to the cost of battery packs and the price of gasoline. When workplace charging is available, the optimal electric driving range surprisingly increases from 16 to 22 miles, as larger batteries would allow drivers to better take advantage of the charging opportunities to achieve longer electrified travel distances, yielding social cost savings. If workplace charging is available, the optimal density is to deploy a workplace charger for every 3.66 vehicles. Moreover, the diversification of the battery size, i.e., introducing a pair and triple of electric driving ranges to the market, could further decrease the average societal cost per PHEV by 7.45% and 11.5% respectively.     

Second order macroscopic traffic flow model validation using automatic differentiation with resilient backpropagation and particle swarm optimisation algorithms

The problem of validating the Modéle d’Écoulement de Trafic sur Autoroute NETworks (METANET) model of a motorway section is considered. Model calibration is formulated as a least squares error minimisation problem with explicit penalisation of fundamental diagram parameter variation. The Automatic Differentiation by Overloading in C++ (ADOL-C) library is incorporated into the METANET source code and is coupled with the Resilient Back Propagation (RPROP) heuristic for solving the minimisation problem. The result is a very efficient system which is able to be calibrate METANET by determining the density and speed equation parameters as well as the fundamental diagrams used. Information obtained from the system’s Jacobian provides extra insight into the dynamics showing how sensitivities propagate into the network. A 22 km site near Sheffield, UK, using data from three different days is considered. In addition to the ADOL-C/RPROP system, three particle swarm optimisation algorithms are used for solving the calibration problem. In all cases, the optimal parameter sets found are verified on data not used during calibration. Although, all three sets of data display a similar congestion pattern, the verification process showed that only one of them is capable of leading to parameter sets that capture the underlying dynamics of the traffic flow process.     

Optimizing route choice for lowest fuel consumption – Potential effects of a new driver support tool

Today, driver support tools intended to increase traffic safety, provide the driver with convenient information and guidance, or save time are becoming more common. However, few systems have the primary aim of reducing the environmental effects of driving. The aim of this project was to estimate the potential for reducing fuel consumption and thus the emission of CO₂ through a navigation system where optimization of route choice is based on the lowest total fuel consumption (instead of the traditional shortest time or distance), further the supplementary effect if such navigation support could take into account real-time information about traffic disturbance events from probe vehicles running in the street network. The analysis was based on a large database of real traffic driving patterns connected to the street network in the city of Lund, Sweden. Based on 15 437 cases, the fuel consumption factor for 22 street classes, at peak and off-peak hours, was estimated for three types of cars using two mechanistic emission models. Each segment in the street network was, on a digitized map, attributed an average fuel consumption for peak and off-peak hours based on its street class and traffic flow conditions. To evaluate the potential of a fuel-saving navigation system the routes of 109 real journeys longer than 5 min were extracted from the database. Using Esri’s external program ArcGIS, Arcview and the external module Network Analysis, the most fuel-economic route was extracted and compared with the original route, as well as routes extracted from criterions concerning shortest time and shortest distance. The potential for further benefit when the system employed real-time data concerning the traffic situation through 120 virtual probe vehicles running in the street network was also examined. It was found that for 46% of trips in Lund the drivers spontaneous choice of route was not the most fuel-efficient. These trips could save, on average, 8.2% fuel by using a fuel-optimized navigation system. This corresponds to a 4% fuel reduction for all journeys in Lund. Concerning the potential for real-time information from probe vehicles, it was found that the frequency of disturbed segments in Lund was very low, and thus so was the potential fuel-saving. However, a methodology is presented that structures the steps required in analyzing such a system. It is concluded that real-time traffic information has the potential for fuel-saving in more congested areas if a sufficiently large proportion of the disturbance events can be identified and reported in real-time.     

A novel passenger flow prediction model using deep learning methods

Currently, deep learning has been successfully applied in many fields and achieved amazing results. Meanwhile, big data has revolutionized the transportation industry over the past several years. These two hot topics have inspired us to reconsider the traditional issue of passenger flow prediction. As a special structure of deep neural network (DNN), an autoencoder can deeply and abstractly extract the nonlinear features embedded in the input without any labels. By exploiting its remarkable capabilities, a novel hourly passenger flow prediction model using deep learning methods is proposed in this paper. Temporal features including the day of a week, the hour of a day, and holidays, the scenario features including inbound and outbound, and tickets and cards, and the passenger flow features including the previous average passenger flow and real-time passenger flow, are defined as the input features. These features are combined and trained as different stacked autoencoders (SAE) in the first stage. Then, the pre-trained SAE are further used to initialize the supervised DNN with the real-time passenger flow as the label data in the second stage. The hybrid model (SAE-DNN) is applied and evaluated with a case study of passenger flow prediction for four bus rapid transit (BRT) stations of Xiamen in the third stage. The experimental results show that the proposed method has the capability to provide a more accurate and universal passenger flow prediction model for different BRT stations with different passenger flow profiles.     

The impact of urban rail transit on commercial property value: New evidence from Wuhan,China

The interaction between rail transit and the urban property market is a vital foundation for planning transit-based policy such as Value Capture and Transit Oriented Development (TOD). Yet only few studies have reported the impact of transit access on commercial property value. This paper presents empirical evidence from Wuhan, China, to enrich the knowledge in the subject area. Spatial autoregressive models were employed to estimate the commercial value capture, based on 676 observations along Wuhan’s metro rail line through the main business districts. Value appreciation was discovered within the 400 m radius of road network distance from Metro stations. The transit access premiums present as two tiers: 16.7% for the 0–100 m core area and approximately 8.0% within the 100–400 m radius. The result demonstrates the potential benefit of adopting value capture and optimising TOD planning to support sustainable urban rail transit investment. Amid rapid urbanisation in China, the evidence reported here could help better inform cities, across the developing world and beyond, of the benefits of adopting rail transit-based policy.