Route travel time estimation using low-frequency floating car data

Assessing the Accuracy of Average Travel Speeds from Floating Car Data on Urban Street Facilities

With advances in information technologies, floating car data (FCD) (e.g. trajectories of taxis) becomes an important data source for collecting traffic information in urban road networks. Due to the low sampling frequency of FCD, it is a challenge to obtain accurate and reliable travel time estimation based on FCD. This paper presents a location-speed method for accurately estimating travel times in signalized road networks based on FCD. The proposed model utilizes trajectories of probe vehicles on the upstream and downstream of signalized intersections in order to estimate delays caused by traffic signals. A large-scale case study using real-world FCD in Wuhan is carried to validate the proposed method. The results of case study show that the proposed method can obtain satisfied estimation of travel times in signalized road networks.

Commercial floating car data application in Sub-Saharan African transport planning contexts: A critical review and research agenda

This paper presents a Real-time Traveler Information System (RTIS) for Hong Kong, in which a novel solution algorithm is proposed for estimation of both the current and instantaneous travel times using automatic vehicle identification (AVI) data. The proposed algorithm also deduces the travel times on the other road links without real-time AVI data. By integration of the filtered real-time and off-line database together, the RTIS can provide area-wide traffic information in the whole network of Hong Kong. The travel time estimates are updated once every five minutes in the RTIS. Observation surveys are conducted during different time periods at a selected path in Hong Kong urban area to validate the RTIS travel time estimates. With the use of the same set of observation data, comparison is also given to the estimation results of the other three existing AVI travel time estimation algorithms: TransGuide, TranStar and Transmit algorithms in US. It was found that the proposed RTIS algorithm could generate the current travel time estimates for the three study periods with the minimum absolute errors and absolute percentage errors. The estimations of instantaneous travel times on the road sections with and without real-time AVI data have also been illustrated. The validation results show that the performance of RTIS is satisfactory and acceptable.

Route travel time variability estimates provide performance information related to the reliability of a trip and allow for confidence bands to be placed around the mean travel time estimates. The naïve (and sometimes used) method to estimate route travel time variance (reliability) is to assume independence of link travel times and consequently sum the individual link variances along the route. In this approach, correlation between links is assumed as zero and the approach is straightforward, but assuming independence is not realistic. This paper describes a post-processing procedure for providing improved route travel time mean and variance estimates, while taking into consideration the correlation existent between individual link travel times. Using automatic vehicle identification (AVI) and inductance loop detector (ILD) data from two separate routes in Texas, a practical application of the theory established by Fu and Rilett (1998. Expected shortest paths in dynamic and stochastic traffic networks. Transportation Research Part B: Methodological, 32(7), 499–516) is used to quantify link travel time correlation. The paper also provides an investigation on the usefulness of the loess statistical method and a polynomial regression model to estimate the distributional properties of link and route travel times. Another significant finding is that route reliability estimated from link ILD speeds (extrapolated to travel times) was not correlated to actual route travel time reliability measured by simultaneously operated probe vehicles. This work is unique in that instrumented probe vehicles were operated at exactly the same times as the roadway sensor data were collected, allowing direct comparison of the travel time estimates from all empirical data sources. The research presents valuable insight on how confidence intervals may be placed on travel time mean estimates for all traffic conditions. With the increased use of travel time data sources such as smartphones, connected vehicles, and private-sector data sources, the methods presented in this paper are invaluable for effective transportation system performance monitoring of both persons and freight movement.

Method of floating car data (FCD) is effective for traffic state estimation. However, using only position and time information sent by FCD vehicles, for short road segments with long cycle time, the averaged FCD travel time could not reflect real traffic state due to the sample size limitation. A data preprocess method is proposed aiming at improving estimation accuracy at signalized road segments using existing sample data and effective green ratios. It is shown both in theoretical proofs and empirical results that the method is effective and practical.

As traffic demand continues to grow in many densely populated urban areas around the world, traffic congestion has become an increasingly critical challenge for transportation agencies to deal with, causing travel delays and environmental pollution. Adaptive traffic signal control systems based on data from associated sensors are used to address the problem. However, often, sensors have errors, such as undercounting actual flows. This dissertation develops an optimization-based methodology, by which undercounted flows by inductive loop detectors can be corrected, and link level travel times can be estimated. The proposed models are validated by both a simulation environment and a real-world case study. Through day-by-day training, the performance of the models improves and stabilizes.The approach is based on a physics-based Traffic Queue model, as a general method for travel time estimation for urban links with typical layouts. A negative binomial regression-based gap inference model is proposed to correct the undercounting error of inductive loop detectors (ILD). The model estimates the number of missing gaps within a time window of one minute. Without ground truth flow count data to train the gap inference model, an optimization platform is proposed to train the model parameters in an indirect way. The model minimizes overall error with respect to link level travel time and flow conservation of intersections. Travel time data are obtained from API such as Google. The proposed approach is validated and illustrated using both simulated and actual data from London, UK. London uses the adaptive control system SCOOT for traffic management. SCOOT ILDs are known to undercount flows under congested conditions. The results show that the model can estimate travel times with good accuracy under heavy congestion, and can tolerate the flow count error up to around 30%. The location-specific model parameters, trained and applied in a day-by-day scenario, show a stabilizing trend and the travel time estimation for online applications improves over time.--Author's abstract

Modern design concepts apply probability density functions to model the traffic flow and capacity. In the application travel times are calculated by drawing random realisations or by estimating an expected value. Generally it is assumed that the distributions of different infrastructure elements are independent and not correlated. In literature this property is so far not often tested. During a project in November 2003 at IVT data was collected from floating cars using a device to record GPS-data on approximately 30,000 kilometres within the canton of Zurich. These floating car data (FCD) was aggregated to gain link travel times on the network of that area. In this paper the correlation of the link travel times or more generally the experienced travel speed is computed, grouped within different road types. This can be done by matching the GPS data to a network model. Different network dependencies are investigated by calculating the correlation for travel speeds measured on paths for given numbers of intersections and distances. It can be seen that the correlation is high for pairs of speeds which are near to each other and decreasing with an increase in the distance. The FCD of three vehicles over many days is combined by establishing a realistic space-time envelope based on theoretically possible routes to the position of other measurements. From each link with an aggregated travel time the FCD the shortest path (based on the travel time) to each later link information was calculated. If these two measurements on the links could have been produced by a car travelling that route at the time this link pair was also used for the calculation of the correlation. With these theoretical routes sufficient data pairs can be compared to get significant values for the correlation.

Probabilistic travel time progression and its application to automatic vehicle identification data

The objective of this paper is to evaluate travel time estimation techniques for off-line and on-line applications using traffic speed. The study area was the 8.1-km Chalerm Mahanakorn Expressway Daokanong - Port section in Bangkok. Traffic speeds were collected at 7 stations. Average traffic volume of 6,140 vph (vehicles per hour) was observed during the 6:00am to 22:00pm period. Four offline methods were used to convert speed into travel time for each link, namely: mid-point, section average, weighted section average, and San Antonio's (lower speed). To aggregate link travel time into route travel time, the instantaneous and time slice methods were considered. The estimated travel times were then compared with the actual travel time observed from 1,632 vehicles. The results indicate that the accuracy depends on the time period and level of traffic congestion. Considering route travel time, the time slice method gives a slightly better accuracy than the instantaneous method. Comparing the offline (all data present) and online (short-term forecast of speed in the next time slice(s)) travel time estimation, the two methods yield similar accuracy. The findings of the research imply that the practical travel time estimation may need several methods suitable for each time period. The improved estimation can increase the accuracy over a single method, thereby decreasing the MAPE (mean absolute percentage error) from 14-20% to 11%.

Abstract. Traffic congestion is a dynamic spatial and temporal process and as such might not be possible to model with linear functions of various dependent variables. That leaves a lot of space for non-linear approximates, such as neutral networks and fuzzy logic. In this paper, the focus is on the fuzzy logic as a possible approach for dealing with the problems of measuring traffic congestion. We investigate the application of this framework on a selected case study, and use floating car data (FCD) collected in Augsburg, Germany. A fuzzy inference system is built to detect degrees of congestion on a federal highway B17. With FCD, it is possible to obtain local speed information on almost all parts of the network. This information, together with collected vehicle location, time and heading, can be further processed and transformed into valuable information in the form of trip routes, travel times, etc. Initial results are compared with traditional method of expressing levels of congestion on a road network e.g. Level of Service – LOS. The fuzzy model, with segmented mean speed and travel time parameters, performed well and showed to be promising approach to detect traffic congestions. This approach can be further improved by involving more input parameters, such as density or vehicle flow, which might reflect traffic congestion event even more realistically.

Travel time is an important parameter in evaluating the operating efficiency of traffic networks, in assessing the performance of traffic management strategies, and as input to many intelligent transportation systems applications such as advanced traveler information systems. Travel time can be obtained directly from instrumented test vehicles, license plate matching, probe vehicles etc., or from indirect methods such as inductance loop detectors. Because of the widespread deployment of loop detectors, they are one of the most widely used inputs to travel time estimation techniques. There are different methods available to calculate the travel time from loop detector data, such as extrapolation of the point speed values, statistical methods, and models based on traffic flow theory. However, most of these methods fail during the transition period between the normal and congested flow conditions. The present study proposes several modifications to an existing traffic flow theory based model for travel time estimation on freeways, such that the model can estimate travel time for varying traffic flow conditions, including transition period, directly from the loop detector data. Field data collected from the I-35 freeway in San Antonio, Tex., USA, are used for illustrating the results. Automatic vehicle identification data collected from the same location are used for validating the results. Simulated data using CORSIM simulation software are also used for the validation of the model.

On urban arterials, travel time estimation is challenging especially from various data sources. Typically, fusing loop detector data and probe vehicle data to estimate travel time is a troublesome issue while considering the data issue of uncertain, imprecise and even conflicting. In this paper, we propose an improved data fusing methodology for link travel time estimation. Link travel times are simultaneously pre-estimated using loop detector data and probe vehicle data, based on which Bayesian fusion is then applied to fuse the estimated travel times. Next, Iterative Bayesian estimation is proposed to improve Bayesian fusion by incorporating two strategies: 1) substitution strategy which replaces the lower accurate travel time estimation from one sensor with the current fused travel time; and 2) specially-designed conditions for convergence which restrict the estimated travel time in a reasonable range. The estimation results show that, the proposed method outperforms probe vehicle data based method, loop detector based method and single Bayesian fusion, and the mean absolute percentage error is reduced to 4.8%. Additionally, iterative Bayesian estimation performs better for lighter traffic flows when the variability of travel time is practically higher than other periods.

Real-time traffic status information provides good references for urban traffic control and management. Travel time is easy to understand and widely employed in representing traffic status. With significantly improved positioning accuracy and coverage, trajectory data collected from GPS-equipped probe vehicles have great potential for traffic state recognition. This paper presents a machine learning enabled travel time estimation method based on the GPS-equipped probe vehicles data. This research considers the spatial-temporal relevancy while solving the travel time allocation problem: the travel time of target segment might be associated with its previous travel times and/or the traffic states of nearby relevant segments. After data normalization and network clustering, an artificial neural network (ANN) algorithm considering such spatial-temporal relevancy was conducted to infer the travel time distribution among the traveled segments within one path. Furthermore, a weighted summation of the travel time estimation result from various trajectories was calculated to better represent the segment travel time in one time step. The proposed method was evaluated by evaluating the estimation results with automatic vehicle identification obtained ground truth. The experimental results illustrated that by utilizing the ANN to consider the spatial-temporal relevancy, the proposed method is effective and efficient in estimating the travel time.

The accuracy of travel time information disseminated in real-time on metropolitan freeways is one of the key issues in the development of advanced traveler information systems. It is generally considered that travel time estimations based on spot speed measurements at loop detectors are not accurate enough to support this real-time information need. This has brought traffic agencies to the view that for real-time information systems to be effective they would need to have much more accurate ways of measuring travel times. Many fancy technologies to directly measure vehicular travel times are being proposed. This chapter shows that, in the real-time context, the precision of the system is not related solely to the accuracy of the measurement. Immediacy and forecasting capabilities play a role. Therefore, focusing only on the accuracy of the travel time measurement is a myopic approach, which can lead to counterintuitive results. Specifically, it is claimed that using travel times estimated with the traditional spot speed Midpoint Algorithm, the performance of the real-time information system is better than by using much more accurate directly measured travel times. Guidelines for an adequate configuration of the common parameters of the system are provided. In addition, real-time context enhancements for travel time estimation methods based on punctual speed measurements are proposed. These are addressed by taking into account an easy and practical implementation. They have been proven to work well in an empirical application on a Spanish Freeway.