Arterial travel time forecast with streaming data: A hybrid approach of flow modeling and machine learning

Article · November 2012with104 Reads
DOI: 10.1016/j.trb.2012.03.006
This article presents a hybrid modeling framework for estimating and predicting arterial traffic conditions using streaming GPS probe data. The model is based on a well-established theory of traffic flow through signalized intersections and is combined with a machine learning framework to both learn static parameters of the roadways (such as free flow velocity or traffic signal parameters) as well as to estimate and predict travel times through the arterial network. The machine learning component of the approach uses the significant amount of historical data collected by the Mobile Millennium system since March 2009 with over 500 probe vehicles reporting their position once per minute in San Francisco, CA.
    • Their relationship is q = ρv. We make the assumption of a triangular FD (fundamental diagram) (Figure 2a), which is also used for arterial traffic estimation and control in the literature (Ban et al., 2011;Liu et al., 2009;Geroliminis & Skabardonis, 2010;Hofleitner, Herring, & Bayen, 2012). The triangular fundamental diagram is determined by several parameters: q max , the capacity (veh/s), v f , the free flow speed (m/s), and ρ max , the jam density (veh/m).
    [Show abstract] [Hide abstract] ABSTRACT: Autonomous driving has become a popular topic in both industry and academia. Lane-changing is a vital component of autonomous driving behavior in arterial road traffic. Much research has been carried out to investigate discretionary lane changes for autonomous vehicles. However, very little research has been conducted on assisting autonomous vehicles in making mandatory lane changes (MLCs), which is the core of optimal lane-specific route planning for autonomous vehicles. This research aims to determine the best position for providing MLC instruction to autonomous vehicles. In this article, an optimization model is formulated to determine the optimal position at which an instruction to change lanes should be given through automotive navigation systems. First, the distribution of time spent waiting for safe headway to make a lane change is modeled as an exponential distribution. Lane-specific travel times are then calculated for vehicles in various situations by applying traffic shockwave theory and horizontal queuing theory. Finally, the expected travel time is derived for a vehicle receiving a lane change instruction to change lanes at an arbitrary position along the road. The proposed model is validated by a comparison with a simulation model in VISSIM. Additional experiments show that the instruction should be given earlier in the case of denser traffic or higher travel speed in the target lane and that vehicles can save considerable time, if they follow the guidance provided by the proposed model. The proposed model can be applied to guide autonomous vehicles to travel an optimal route.
    Full-text · Article · Apr 2017
    • Throughout the analyzed literature, several research projects have been referred and sometimes its datasets have been used to numerically evaluate the techniques suggested: @BULLET " Mobile Millennium " [13], [10], [9], [21] – developed by the California Center for Innovative Transportation (CCIT), the Nokia Research Center (NRC), and the University of California (UC) at Berkeley [46]. The goal of the project was to the explore the capability of GPSenabled mobile phones to provide traffic data for the purpose of estimating real-time conditions and forecast future conditions [21].
    [Show abstract] [Hide abstract] ABSTRACT: Automatic Vehicle Location (AVL) is becoming an important tool in Intelligent Transportation Systems (ITS) in the past few years, as it is an effective way of collecting and transmitting data regarding the vehicle's trip for real-time or future use. A methodology for analyzing the state of the art regarding the application of these systems is proposed in a form of a systematic literature review, by identifying and systematizing possible transportation network performance metrics that can be calculated or predicted using GPS-based AVL systems and inferring tendencies observed throughout the literature regarding techniques used and sensor data source and usage. As a result of this research, several performance metrics were identified, with Travel Time and Average Speed being the most recurrent ones. The conclusions reveal an increase in the number of publications and research projects regarding this topic over the years, as well as a promising potential of this type of technology, with buses and taxis being the most used probe vehicles.
    Full-text · Conference Paper · Nov 2016 · Transportation Research Procedia
    • Thus, short-term forecasting queue lengths at ferry terminals has important practical significance to both traffic service management interests and the travelers in cities along waterfront areas. For traffic forecasting, typical methods include statistical time series analysis methods [5, 6], machine learning techniques [7], fuzzy logic [8], Bayesian networks [9], and other prediction models developed with specific applications for ferry traffic [10], among many others. Recent studies have proposed hybrid methods that accommodate the cyclical pattern of traffic data [11].
    [Show abstract] [Hide abstract] ABSTRACT: Ferry service plays an important role in several cities with waterfront areas. Transportation authorities often need to forecast volumes of vehicular traffic in queues waiting to board ships at ferry terminals to ensure sufficient capacity and establish schedules that meet demand. Several previous studies have developed models for long-term vehicle queue length prediction at ferry terminals using terminal operation data. Few studies, however, have been undertaken for short-term vehicular queue length prediction. In this study, machine learning methods including the artificial neural network (ANN) and support vector machine (SVM) are applied to predict vehicle waiting queue lengths at ferry terminals. Through time series analysis, the existence of a periodic queue-length pattern is established. Hence, methodologies used in this study take into account periodic features of vehicle queue data at terminals for prediction. To further consider the cyclical characteristics of vehicle queue data at ferry terminals, a prediction approach is proposed to decompose vehicle waiting queue length into two components: a periodic part and a dynamic part. A trigonometric regression function is introduced to capture the periodic component, and the dynamic part is modeled by SVM and ANN models. Moreover, an assembly technique for combining SVM and ANN models is proposed to aggregate multiple prediction models and in turn achieve better results than could be attained from a lone predictive method. The prediction results suggest that for multi-step ahead vehicle queue length prediction at ferry terminals, the ensemble model outperforms the separate prediction models and the hybrid models, especially as prediction step size increases. This research has important practical significance to both traffic service management interests and the travelers in cities along waterfront areas.
    Full-text · Article · May 2016
    • We use our method to predict ambulance travel times for the entire road network of Toronto. The size of the road network (68,272 links) is an order of magnitude larger than in previous applications of travel time distribution estimation based on floating car data [6, 14, 15, 17, 37], and the number of historical vehicle trips (157,283) is also larger than these previous applications. We compare the prediction accuracy of our method to that of Budge et al. [6], Westgate et al. [37], and a commercial software package for mean travel time estimation.
    [Show abstract] [Hide abstract] ABSTRACT: We propose a regression approach for estimating the distribution of ambulance travel times between any two locations in a road network. Our method uses ambulance location data that can be sparse in both time and network coverage, such as Global Positioning System data. Estimates depend on the path traveled and on explanatory variables such as the time of day and day of week. By modeling at the trip level, we account for dependence between travel times on individual road segments. Our method is parsimonious and computationally tractable for large road networks. We apply our method to estimate ambulance travel time distributions in Toronto, providing improved estimates compared to a recently published method and a commercial software package. We also demonstrate our method’s impact on ambulance fleet management decisions, showing substantial differences between our method and the recently published method in the predicted probability that an ambulance arrives within a time threshold.
    Full-text · Article · Jan 2016
    • Likewise, studies have proposed forecasting bus travel times using KF (Wall and Dailey, 1999; Chen et al., 2012, Hans et al., 2014b). Highly non-linear models and stochastic systems require a particle filter (PF) to easily compute all possible system states and so provides the most likely state (Hofleitner et al., 2012). Such data assimilation methods can be particularly efficient for predicting future travel times (Chen and Rakha, 2014).
    [Show abstract] [Hide abstract] ABSTRACT: Buses on the same route tend to bunch when the system is uncontrolled. This lack of regularity leads to an increase in the average passenger waiting time, increases delays and makes travel times uncertain. A wide variety of solutions have been proposed to maintain accurate bus system performance. Unfortunately, if a strategy is applied permanently, it could detract from the entire transport system efficiency. That is why a transit operator needs an accurate forecast of the route in order to intervene before the bus route is too disrupted to be restored to regularity. This paper aims to predict critical situations in real-time forecasting of a bus route state. To accomplish this, we propose to take advantage of both theoretical and empirical information (model and data) using data assimilation (a particle filter). On one hand, a stochastic dynamic bus model forecasts future bus route states. On the other hand, archived data calibrates the model parameters while real-time data provides information about the actual route state. The methodology is applied to a real case study thanks to the quality data provided by TriMet (the Portland, Oregon transit district). Predictions are finally evaluated by an a posteriori comparison with real data. The results highlight that the method leads to a valid forecast of a bus route state with a 8 minutes time window. This duration is sufficient to predict critical situations, especially bus bunching. Further research would have to consider deterministic travel times from a traffic model instead of the distributions in order to maintain correlation between travel times on links. In that case, the assimilation process would focus on the surrounding traffic flow, also potentially available in the Portland data.
    Full-text · Article · Dec 2015
    • To estimate travel times, methods can be divided into two types: 1. Directly estimate the travel time with samples using various statistical models, such as the " seemingly unrelated regression equations " (SURE) method (Martchouk et al., 2011 ), Student's t distribution estimation (Haghani et al., 2010; Quayle et al., 2010; Richardson et al., 2011), maximum likelihood estimation (Kwong et al., 2009; Sanchez et al., 2011a), the least quartile of squares method (Bhaskar et al., 2011; Van Boxel et al., 2011), multiple variants of neural networks (Van Lint and Hoogendoorn, 2010; Zeng and Zhang, 2013; Zhang and Ge, 2013). 2. Bayesian estimation using Kalman Filtering (Qiu et al., 2009; Barcelo et al., 2010) or Dynamic Bayesian Networks (Hofleitner et al., 2012).
    File · Data · Aug 2015 · Transportation Research Procedia
Show more
December 2012 · IEEE Transactions on Intelligent Transportation Systems · Impact Factor: 2.38
    Estimating and predicting traffic conditions in arterial networks using probe data has proven to be a substantial challenge. Sparse probe data represent the vast majority of the data available on arterial roads. This paper proposes a probabilistic modeling framework for estimating and predicting arterial travel-time distributions using sparsely observed probe vehicles. We introduce a model... [Show full abstract]
    Conference Paper
    October 2010 · Conference Record - IEEE Conference on Intelligent Transportation Systems · Impact Factor: 0.25
      Estimating and predicting traffic conditions in arterial networks using probe data has proven to be a substantial challenge. In the United States, sparse probe data represents the vast majority of the data available on arterial roads in most major urban environments. This article proposes a probabilistic modeling framework for estimating and predicting arterial travel time distributions using... [Show full abstract]
        In arterial networks, traffic flow dynamics are driven by the presence of traffic signals, for which precise signal timing is difficult to obtain in arbitrary networks or might change over time. A comprehensive model of arterial traffic flow dynamics is necessary to capture its specific fea-tures in order to provide accurate traffic estimation approaches. From hydrodynamic theory, we model... [Show full abstract]
        January 2009
          The mobile internet is changing the face of the transportation cyberphysical system at a rapid pace. In the last five years, cellular phone technology has leapfrogged several attempts to construct dedicated infrastructure systems to monitor traffic. Today, GPS equipped smartphones are progressively morphing into a ubiquitous traffic monitoring system where users contribute and receive traffic... [Show full abstract]
          Discover more