Online Prediction of Travel Time: Experience From a Pilot Trial

Abstract

This study was designed to present an online model which predicted travel times on an interurban two-lane two-way highway section on the basis of field measurements. The study included two parts: an evaluation of the performance of the model, and an examination of the possibility to improve the model in case of unsatisfactory performance. The model was based on MLP neural networks. The main results of the evaluation showed that the prediction model outperformed a non-predictive system. However, the model for one section had not performed as well during the trial period as was expected. This might be due to a slight change in the congestion phenomenon. After further development, the findings showed that the model could be improved considerably with new data. The main implication was that even a simple prediction model improves the quality of travel time information substantially, compared to estimates based directly on the latest measurements.     

Incorporating uncertainty into short-term travel time predictions

To improve the quality of travel time information provided to motorists, there is a need to move away from point forecasts of travel time. Specifically, techniques are needed which predict the range of travel times which motorists may experience. This paper focuses on travel time prediction on motorways and evaluates three models for predicting the travel time range in real time as well as up to 1 h ahead. The first model, termed lane by lane tracing, relies on speed data from each lane to replicate the trajectories of relatively slow and relatively fast vehicles on the basis of speed differences across the lanes. The second model is based on the relationship between mean travel time (estimated using a neural network model) and driver-to-driver travel time variability. The results provide insight into the relative merits of the proposed techniques and confirm that they provide a basis for reliable travel time range prediction in the short-term prediction context (up to 1 h ahead).     

Information provision strategies eliminating deluded equilibrium caused by travellers’ misperception

Providing travel time information may be effective at reducing travel costs. However, this information does not always match the actual travel time that travellers will experience. Furthermore, the information is often asymmetrically provided within the network, owing to the limitations of observation devices, prediction model calibration, and uncertainty about road conditions. The purpose of this study is to investigate the effects of predictive travel time information that is asymmetrically provided to travellers. This study formulated a dynamic traffic assignment model in origin–destination (OD) pair with two parallel routes, while considering travellers’ learning processes and within-day and day-to-day dynamics. In this study, it is assumed that different information will be provided to each traveller, according to within-day traffic dynamics. Furthermore, the information is provided for only one of two possible routes, because of observation limitations. The effects of information accuracy are also discussed in this study. The results of numerical analysis indicated that information provisions possibly reduced the negative effects of deluded equilibrium state, even when the information was only provided for one of the routes. Different effects of the travel time information and its variation were illustrated according to the allocation of the bottleneck capacities of two routes.     

Predicting travel time reliability using mobile phone GPS data

Estimates of road speeds have become commonplace and central to route planning, but few systems in production provide information about the reliability of the prediction. Probabilistic forecasts of travel time capture reliability and can be used for risk-averse routing, for reporting travel time reliability to a user, or as a component of fleet vehicle decision-support systems. Many of these uses (such as those for mapping services like Bing or Google Maps) require predictions for routes in the road network, at arbitrary times; the highest-volume source of data for this purpose is GPS data from mobile phones. We introduce a method (TRIP) to predict the probability distribution of travel time on an arbitrary route in a road network at an arbitrary time, using GPS data from mobile phones or other probe vehicles. TRIP captures weekly cycles in congestion levels, gives informed predictions for parts of the road network with little data, and is computationally efficient, even for very large road networks and datasets. We apply TRIP to predict travel time on the road network of the Seattle metropolitan region, based on large volumes of GPS data from Windows phones. TRIP provides improved interval predictions (forecast ranges for travel time) relative to Microsoft’s engine for travel time prediction as used in Bing Maps. It also provides deterministic predictions that are as accurate as Bing Maps predictions, despite using fewer explanatory variables, and differing from the observed travel times by only 10.1% on average over 35,190 test trips. To our knowledge TRIP is the first method to provide accurate predictions of travel time reliability for complete, large-scale road networks.     

Forecasting journey time distribution with consideration to abnormal traffic conditions

Travel time is an important index for managers to evaluate the performance of transportation systems and an intuitive measure for travelers to choose routes and departure times. An important part of the literature focuses on predicting instantaneous travel time under recurrent traffic conditions to disseminate traffic information. However, accurate travel time prediction is important for assessing the effects of abnormal traffic conditions and helping travelers make reliable travel decisions under such conditions. This study proposes an online travel time prediction model with emphasis on capturing the effects of anomalies. The model divides a path into short links. A Functional Principal Component Analysis (FPCA) framework is adopted to forecast link travel times based on historical data and real-time measurements. Furthermore, a probabilistic nested delay operator is used to calculate path travel time distributions. To ensure that the algorithm is fast enough for online applications, parallel computation architecture is introduced to overcome the computational burden of the FPCA. Finally, a rolling horizon structure is applied to online travel time prediction. Empirical results for Guangzhou Airport Expressway indicate that the proposed method can capture an abrupt change in traffic state and provide a promising and reliable travel time prediction at both the link and path levels. In the case where the original FPCA is modified for parallelization, accuracy and computational effort are evaluated and compared with those of the sequential algorithm. The proposed algorithm is found to require only a piece rather than a large set of traffic incident records.     

A bayesian dynamic linear model approach for real-time short-term freeway travel time prediction

This paper presents a Bayesian inference-based dynamic linear model (DLM) to predict online short-term travel time on a freeway stretch. The proposed method considers the predicted freeway travel time as the sum of the median of historical travel times, time-varying random variations in travel time, and a model evolution error, where the median is employed to recognize the primary travel time pattern while the variation captures unexpected supply (i.e. capacity) reduction and demand fluctuations. Bayesian forecasting is a learning process that revises sequentially the state of a priori knowledge of travel time based on newly available information. The prediction result is a posterior travel time distribution that can be employed to generate a single-value (typically but not necessarily the mean) travel time as well as a confidence interval representing the uncertainty of travel time prediction. To better track travel time fluctuations during non-recurrent congestion due to unforeseen events (e.g., incidents, accidents, or bad weather), the DLM is integrated into an adaptive control framework that can automatically learn and adjust the system evolution noise level. The experiment results based on the real loop detector data of an I-66 segment in Northern Virginia suggest that the proposed method is able to provide accurate and reliable travel time prediction under both recurrent and non-recurrent traffic conditions.     

Probe vehicle based real-time traffic monitoring on urban roadways

Travel time estimation and prediction on urban arterials is an important component of Active Traffic and Demand Management Systems (ATDMS). This paper aims in using the information of GPS probes to augment less dynamic but available information describing arterial travel times. The direction followed in this paper chooses a cooperative approach in travel time estimation using static information describing arterial geometry and signal timing, semi-dynamic information of historical travel time distributions per time of day, and utilizes GPS probe information to augment and improve the latter. First, arterial travel times are classified by identifying different travel time states, then link travel time distributions are approximated using mixtures of normal distributions. If prior travel time data is available, travel time distributions can be estimated empirically. Otherwise, travel time distribution can be estimated based on signal timing and arterial geometry. Real-time GPS travel time data is then used to identify the current traffic condition based on Bayes Theorem. Moreover, these GPS data can also be used to update the parameters of the travel time distributions using a Bayesian update. The iterative update process makes the posterior distributions more and more accurate. Finally, two comprehensive case studies using the NGSIM Peachtree Street dataset, and GPS data of Washington Avenue in Minneapolis, were conducted. The first case study estimated prior travel time distributions based on signal timing and arterial geometry under different traffic conditions. Travel time data were classified and corresponding distributions were updated. In addition, results from the Bayesian update and EM algorithm were compared. The second case study first tested the methodologies based on real GPS data and showed the importance of sample size. In addition, a methodology was proposed to distinguish new traffic conditions in the second case study.     

An adaptive information fusion model to predict the short-term link travel time distribution in dynamic traffic networks

As intelligent transportation systems (ITS) approach the realm of widespread deployment, there is an increasing need to robustly capture the variability of link travel time in real-time to generate reliable predictions of real-time traffic conditions. This study proposes an adaptive information fusion model to predict the short-term link travel time distribution by iteratively combining past information on link travel time on the current day with the real-time link travel time information available at discrete time points. The past link travel time information is represented as a discrete distribution. The real-time link travel time is represented as a range, and is characterized using information quality in terms of information accuracy and time delay. A nonlinear programming formulation is used to specify the adaptive information fusion model to update the short-term link travel time distribution by focusing on information quality. The model adapts good information by weighing it higher while shielding the effects of bad information by reducing its weight. Numerical experiments suggest that the proposed model adequately represents the short-term link travel time distribution in terms of accuracy and robustness, while ensuring consistency with ambient traffic flow conditions. Further, they illustrate that the mean of a representative short-term travel time distribution is not necessarily a good tracking indicator of the actual (ground truth) time-dependent travel time on that link. Parametric sensitivity analysis illustrates that information accuracy significantly influences the model, and dominates the effects of time delay and the consistency constraint parameter. The proposed information fusion model bridges key methodological gaps in the ITS deployment context related to information fusion and the need for short-term travel time distributions.     

Bus travel time prediction using a time-space discretization approach

The accuracy of travel time information given to passengers plays a key role in the success of any Advanced Public Transportation Systems (APTS) application. In order to improve the accuracy of such applications, one should carefully develop a prediction method. A majority of the available prediction methods considered the variation in travel time either spatially or temporally. The present study developed a prediction method that considers both temporal and spatial variations in travel time. The conservation of vehicles equation in terms of flow and density was first re-written in terms of speed in the form of a partial differential equation using traffic stream models. Then, the developed speed based equation was discretized using the Godunov scheme and used in the prediction scheme that was based on the Kalman filter. From the results, it was found that the proposed method was able to perform better than historical average, regression, and ANN methods and the methods that considered either temporal or spatial variations alone. Finally, a formulation was developed to check the effect of side roads on prediction accuracy and it was found that the additional requirement in terms of location based data did not result in an appreciable change in the prediction accuracy. This clearly demonstrated that the proposed approach based on using vehicle tracking data is good enough for the considered application of bus travel time prediction.     

Evaluating the added-value of online bus arrival prediction schemes

Online predictions of bus arrival times have the potential to reduce the uncertainty associated with bus operations. By better anticipating future conditions, online predictions can reduce perceived and actual passenger travel times as well as facilitate more proactive decision making by service providers. Even though considerable research efforts were devoted to the development of computationally expensive bus arrival prediction schemes, real-world real-time information (RTI) systems are typically based on very simple prediction rules. This paper narrows down the gap between the state-of-the-art and the state-of-the-practice in generating RTI for public transport systems by evaluating the added-value of schemes that integrate instantaneous data and dwell time predictions. The evaluation considers static information and a commonly deployed scheme as a benchmark. The RTI generation algorithms were applied and analyzed for a trunk bus network in Stockholm, Sweden. The schemes are assessed and compared based on their accuracy, reliability, robustness and potential waiting time savings. The impact of RTI on passengers waiting times are compared with those attained by service frequency and regularity improvements. A method which incorporates information on downstream travel conditions outperforms the commonly deployed scheme, leading to a 25% reduction in the mean absolute error. Furthermore, the incorporation of instantaneous travel times improves the prediction accuracy and reliability, and contributes to more robust predictions. The potential waiting time gains associated with the prediction scheme are equivalent to the gains expected when introducing a 60% increase in service frequency, and are not attainable by service regularity improvements.     

A Time-use Model for the Automated Vehicle-era

Automated Vehicles (AVs) offer their users a possibility to perform new non-driving activities while being on the way. The effects of this opportunity on travel choices and travel demand have mostly been conceptualised and modelled via a reduced penalty associated with (in-vehicle) travel time. This approach invariably leads to a prediction of more car-travel. However, we argue that reductions in the size of the travel time penalty are only a crude proxy for the variety of changes in time-use and travel patterns that are likely to occur at the advent of AVs. For example, performing activities in an AV can save time and in this way enable the execution of other activities within a day. Activities in an AV may also eliminate or generate a need for some other activities and travel. This may lead to an increase, or decrease in travel time, depending on the traveller’s preferences, schedule, and local accessibility. Neglecting these dynamics is likely to bias forecasts of travel demand and travel behaviour in the AV-era. In this paper, we present an optimisation model which rigorously captures the time-use effects of travellers’ ability to perform on-board activities. Using a series of worked out examples, we test the face validity of the model and demonstrate how it can be used to predict travel choices in the AV-era.     

Real-time trip purpose prediction using online location-based search and discovery services

The use of smartphone technology is increasingly considered a state-of-the-art practice in travel data collection. Researchers have investigated various methods to automatically predict trip characteristics based upon locational and other smartphone sensing data. Of the trip characteristics being studied, trip purpose prediction has received relatively less attention. This research develops trip purpose prediction models based upon online location-based search and discovery services (specifically, Google Places API) and a limited set of trip data that are usually available upon the completion of the trip. The models have the potential to be integrated with smartphone technology to produce real-time trip purpose prediction. We use a recent, large-scale travel behavior survey that is augmented by downloaded Google Places information on each trip destination to develop and validate the models. Two statistical and machine learning prediction approaches are used, including nested logit and random forest methods. Both sets of models show that Google Places information is a useful predictor of trip purpose in situations where activity- and person-related information is uncollectable, missing, or unreliable. Even when activity- and person-related information is available, incorporating Google Places information provides incremental improvements in trip purpose prediction.     

Trip generation modeling using data collected in single and repeated cross‐sectional surveys

The majority of US metropolitan regions still use the four‐step urban transportation modeling system to develop their travel forecasts. Trip generation, the first step of this system, has as objective of predicting the expected total travel demand in a region. The commonly used methods in planning practice for predicting this expected total travel demand typically use only the most recent cross‐sectional data available from a study region for model development, which ties the resulting travel‐forecast model to the economic environment prevailing at the time of data collection. Applying such models to generate forecasts of travel in economic environments significantly different from those embodied in the estimated model parameters could result in greater errors than would otherwise be the case. To address the aforementioned problem, this paper proposes the development of trip generation models estimated on multiple independent cross‐sectional datasets collected in the same urban region but at different times representing different economic environments. Data used in the research were collected in cross‐sectional household travel behavior surveys undertaken in the Greater Toronto Area, Canada in 1986, 1996, 2001, and 2006. The results lead to the conclusion that well‐specified models, estimated on pooled multiple cross‐sectional datasets, yield travel predictions in the base and horizon years, respectively, that have smaller error compared with corresponding travel predictions generated with single cross‐sectional models. Copyright © 2012 John Wiley & Sons, Ltd.     

Traveler delay costs and value of time with trip chains, flexible activity scheduling and information

The delay costs of traffic disruptions and congestion and the value of travel time reliability are typically evaluated using single trip scheduling models, which treat the trip in isolation of previous and subsequent trips and activities. In practice, however, when activity scheduling to some extent is flexible, the impact of delay on one trip will depend on the actual and predicted travel time on itself as well as other trips, which is important to consider for long-lasting disturbances and when assessing the value of travel information. In this paper we extend the single trip approach into a two trips chain and activity scheduling model. Preferences are represented as marginal activity utility functions that take scheduling flexibility into account. We analytically derive trip timing optimality conditions, the value of travel time and schedule adjustments in response to travel time increases. We show how the single trip models are special cases of the present model and can be generalized to a setting with trip chains and flexible scheduling. We investigate numerically how the delay cost depends on the delay duration and its distribution on different trips during the day, the accuracy of delay prediction and travel information, and the scheduling flexibility of work hours. The extension of the model framework to more complex schedules is discussed.     

Validation and use of equilibrium network assignment for urban highway reconstruction planning

The need to rehabilitate interstate highways and bridges will increase tremendously in the next decades. Due to traffic restrictions imposed during construction, these rehabilitation activities will cause major disruptions in existing traffic patterns. In order to develop mitigation strategies to reduce such travel impacts, reliable forecasts of likely travel pattern changes would be beneficial. In this paper, we examine the suitability of using an equilibrium traffic assignment model to predict the impacts of a major highway reconstruction project. A case study of travel impacts during reconstruction of I-376, the Parkway East, in Pittsburgh, Pennsylvania is made to validate the adequacy of the network assignment model. Results are compared with actual volume counts collected during periods with and without traffic restrictions. The model produced estimates of link volumes that were, on average, from 16% to 28% different from the observed link counts along two screenlines. Large discrepancies with some of the counts could be explained in part by aberrations in the observed data or in the network model's structure. A sketch-planning analysis is also performed, and the results are compared with those from the network assignment model. The network assignment model is also used to predict the impacts of a hypothetical reconstruction scenario in which the Parkway East is totally closed during its reconstruction.     

Differences in probability weighting for individual travelers: a managed lane choice application

Using a 2012 stated preference survey based on a traveler’s most recent actual trip, this study predicts traveler choices between general purpose lanes and managed lanes for a freeway in Houston, Texas. The choice model incorporates probability weighting for risky travel times. The results indicate significant improvement in predicative power over a model that excludes weighting, confirming non-linearity in the probability weighting function. The maximum value of time (VOT) measures calculated in this study are lower than estimated in many previous route choice studies. This highlights the importance of incorporating individual weights for travel risks. Travelers’ underweighting of travel time risks would help explain the lower VOTs found in our study because respondents consider route choice decision-making as a gamble, but assign their own probabilities of occurrence to arriving at their destination on time, late, or early. We find that traveler groups are heterogeneous and the different weights developed for different groups of travelers can be used to better understand their probabilities. Segmentation analysis indicates that Age may serve to proxy the effects of more experience over time, or changing driving abilities, or changes in one’s sense of optimism or pessimism at different ages. Gender and Income also play a role in how the objective probabilities presented to respondents were translated into subjective probabilities.     

Four-step travel demand model implementation for estimating traffic volumes on rural low-volume roads in Wyoming

This study develops a four-step travel demand model for estimating traffic volumes for low-volume roads in Wyoming. The study utilizes urban travel behavior parameters and processes modified to reflect the rural and low-volume nature of Wyoming local roads. The methodology disaggregates readily available census block data to create transportation analysis zones adequate for estimating traffic on low-volume rural roads. After building an initial model, the predicted and actual traffic volumes are compared to develop a calibration factor for adjusting trip rates. The adjusted model is verified by comparing estimated and actual traffic volumes for 100 roads. The R-square value from fitting predicted to actual traffic volumes is determined to be 74% whereas the Percent Root Mean Square Error is found to be 50.3%. The prediction accuracy for the four-step travel demand model is found to be better than a regression model developed in a previous study.     

Optimal speed detector density for the network with travel time information

In the expressway network, detectors are installed on the links for detecting the travel time information while the predicted travel time can be provided by the route guidance system (RGS). The speed detector density can be determined to influence flow distributions in such a way that the precision of the travel time information and the social cost of the speed detectors are optimized, provided that each driver chooses the minimum perceived travel time path in response to the predicted travel time information. In this paper, a bilevel programming model is proposed for the network with travel time information provided by the RGS. The lower-level problem is a probit-based traffic assignment model, while the upper-level problem is to determine the speed detector density that minimizes the measured travel time error variance as well as the social cost of the speed detectors. The sensitivity analysis based algorithm is proposed for the bilevel programming problem. Numerical examples are provided to illustrate the applications of the proposed model and of the solution algorithm.     

Experienced travel time prediction for congested freeways

Travel time is an important performance measure for transportation systems, and dissemination of travel time information can help travelers make reliable travel decisions such as route choice or departure time. Since the traffic data collected in real time reflects the past or current conditions on the roadway, a predictive travel time methodology should be used to obtain the information to be disseminated. However, an important part of the literature either uses instantaneous travel time assumption, and sums the travel time of roadway segments at the starting time of the trip, or uses statistical forecasting algorithms to predict the future travel time. This study benefits from the available traffic flow fundamentals (e.g. shockwave analysis and bottleneck identification), and makes use of both historical and real time traffic information to provide travel time prediction. The methodological framework of this approach sequentially includes a bottleneck identification algorithm, clustering of traffic data in traffic regimes with similar characteristics, development of stochastic congestion maps for clustered data and an online congestion search algorithm, which combines historical data analysis and real-time data to predict experienced travel times at the starting time of the trip. The experimental results based on the loop detector data on Californian freeways indicate that the proposed method provides promising travel time predictions under varying traffic conditions.