Analysis of variability in multi-day GPS imputed activity-travel diaries using multi-dimensional sequence alignment and panel effects regression models

The use of GPS devices and smartphones has made feasible the collection of multi-day activity-travel diaries. In turn, the availability of multi-day travel diary data opens up new avenues for analyzing dynamics of individual travel behavior. This paper addresses the issue of day-to-day variability in activity-travel behavior. The study, which is the first of its kind in China, applies a unique combination of methods to analyze the degree of dissimilarity between travel days using multi-day GPS data. First, multi-dimensional sequence alignment is applied to measure the degree of dissimilarity in individual daily activity-travel sequences between pairs of travel days. Next, a series of panel effects regression models is used to estimate the effects of socio-demographics and days of the week. The models are estimated using multi-day activity-travel patterns imputed from GPS-enabled smartphone data collected in Shanghai, China. Results indicate that (1) days of the week have significant effects on day-to-day variability in activity-travel behavior with weekday activity-travel sequences being more similar and thereby different from weekend sequences; (2) the degree of dissimilarity in activity-travel sequences is strongly influenced by respondent socio-demographic profiles; (3) individuals having more control over and flexibility in their work schedule show greater intra-personal variability. Day-to-day variability in activity-travel behavior of this sample is similar to patterns observed in developed countries in some aspects but different in others. Strict international comparison study based on comparative data collection is required to further distinguish the sources of travel behavior differences between developing countries and developed countries. The paper ends with a discussion of the limitations of this study and the implications of the research findings for future research.     

Integrated imputation of activity-travel diaries incorporating the measurement of uncertainty

Procedures to transform GPS tracks into activity-travel diaries have been increasingly addressed due to their potential benefit to replace traditional methods used in travel surveys. Existing approaches for data annotation however are not sufficiently accurate, which normally involves a prompted recall survey for data validation. Imputation algorithms for transportation mode detection seem to be largely dependent on speed-related features, which may blur the quality of classification results, especially with transportation modes having similar speeds. Therefore, in this paper we propose an enhanced integrated imputation approach by incorporating the critical indicators related to trip patterns, reflecting the effects of uncertain travel environments, including bus stops and speed percentiles. A two-step procedure which embeds a segmentation model and a transportation mode inference model is designed and examined based on purified prompted recall data collected in a large-scale travel survey. Results show the superior performance of the proposed approach, where the overall accuracy at trip level reaches 93.2% and 88.1% for training and surveyed data, respectively.     

Vehicle classification from low-frequency GPS data with recurrent neural networks

The categorization of the type of vehicles on a road network is typically achieved using external sensors, like weight sensors, or from images captured by surveillance cameras. In this paper, we leverage the nowadays widespread adoption of Global Positioning System (GPS) trackers and investigate the use of sequences of GPS points to recognize the type of vehicle producing them (namely, small-duty, medium-duty and heavy-duty vehicles). The few works which already exploited GPS data for vehicle classification rely on hand-crafted features and traditional machine learning algorithms like Support Vector Machines. In this work, we study how performance can be improved by deploying deep learning methods, which are recently achieving state of the art results in the classification of signals from various domains. In particular, we propose an approach based on Long Short-Term Memory (LSTM) recurrent neural networks that are able to learn effective hierarchical and stateful representations for temporal sequences. We provide several insights on what the network learns when trained with GPS data and contextual information, and report experiments on a very large dataset of GPS tracks, where we show how the proposed model significantly improves upon state-of-the-art results.     

Day-to-day variability in travellers’ activity-travel patterns in the Jakarta metropolitan area

Using four consecutive days of SITRAMP 2004 data from the Jakarta metropolitan area (JMA), Indonesia, this study examines the interactions between individuals’ activity-travel parameters, given the variability in their daily constraints, resources, land use and road network conditions. While there have been a significant number of studies into day-to-day variability in travel behaviour in developed countries, this issue is rarely examined in developing countries. The results show that some activity-travel parameter interactions are similar to those produced by travellers from developed countries, while others differ. Household and individual characteristics are the most significant variables influencing the interactions between activity-travel parameters. Different groups of travellers exhibit different trade-off mechanisms. Further analyses of the stability of activity-travel patterns across different days are also provided. Daily commuting time and regular work and study commitments heavily shape workers’ and students’ flexibility in arranging their travel time and out-of-home time budget, leading to more stable daily activity-travel patterns than non-workers.     

Effective arterial road incident detection: A Bayesian network based algorithm

Timely and accurate incident detection is an essential part of any successful advanced traffic management system. The complex nature of arterial road traffic makes automated incident detection a real challenge. Stable performance and strong transferability remain major issues concerning the existing incident detection algorithms. A new arterial road incident detection algorithm TSC_ar is presented in this paper. In this algorithm, Bayesian networks are used to quantitatively model the causal dependencies between traffic events (e.g. incident) and traffic parameters. Using real time traffic data as evidence, the Bayesian networks update the incident probability at each detection interval through two-way inference. An incident alarm is issued when the estimated incident probability exceeds the predefined decision threshold. The Bayesian networks allow us to subjectively build existing traffic knowledge into their conditional probability tables, which makes the knowledge base for incident detection robust and dynamic. Meanwhile, we incorporate intersection traffic signals into traffic data processing. A total of 40 different types of arterial road incidents are simulated to test the performance of the algorithm. The high detection rate of 88% is obtained while the false alarm rate of the algorithm is maintained as low as 0.62%. Most importantly, it is found that both the detection rate and false alarm rate are not sensitive to the incident decision thresholds. This is the unique feature of the TSC_ar algorithm, which suggests that the Bayesian network approach is advanced in enabling effective arterial road incident detection.     

Short-term speed predictions exploiting big data on large urban road networks

Big data from floating cars supply a frequent, ubiquitous sampling of traffic conditions on the road network and provide great opportunities for enhanced short-term traffic predictions based on real-time information on the whole network. Two network-based machine learning models, a Bayesian network and a neural network, are formulated with a double star framework that reflects time and space correlation among traffic variables and because of its modular structure is suitable for an automatic implementation on large road networks. Among different mono-dimensional time-series models, a seasonal autoregressive moving average model (SARMA) is selected for comparison. The time-series model is also used in a hybrid modeling framework to provide the Bayesian network with an a priori estimation of the predicted speed, which is then corrected exploiting the information collected on other links. A large floating car data set on a sub-area of the road network of Rome is used for validation. To account for the variable accuracy of the speed estimated from floating car data, a new error indicator is introduced that relates accuracy of prediction to accuracy of measure. Validation results highlighted that the spatial architecture of the Bayesian network is advantageous in standard conditions, where a priori knowledge is more significant, while mono-dimensional time series revealed to be more valuable in the few cases of non-recurrent congestion conditions observed in the data set. The results obtained suggested introducing a supervisor framework that selects the most suitable prediction depending on the detected traffic regimes.     

Scheduling decision styles on leisure and social activities

This paper investigates scheduling decisions associated with different types of leisure and social activities. Correlations among decisions and self-selection biases are explicitly investigated by using a sample selection model with a bivariate probit selection rule. A dataset collected in the first wave of a recent activity-travel scheduling panel survey carried out in Valencia (Spain) was used for empirical investigation. Significant differences are revealed in the empirical models for leisure and social activities in planning decisions, including different effects of temporal, companionship and demographic factors. The findings of the empirical model have important implications to travel behavior and activity-travel scheduling model developments. These results confirm the existence of different mechanisms underlying the activity-travel decision processes when leisure and social activities are of concerns. Results provide significant insights into enhancing the performances of an activity scheduling model by capturing accurate activity-travel scheduling tradeoffs in flexible activity types e.g. leisure and social activities.     

A driving cycle detection approach using map service API

Following advancements in smartphone and portable global positioning system (GPS) data collection, wearable GPS data have realized extensive use in transportation surveys and studies. The task of detecting driving cycles (driving or car-mode trajectory segments) from wearable GPS data has been the subject of much research. Specifically, distinguishing driving cycles from other motorized trips (such as taking a bus) is the main research problem in this paper. Many mode detection methods only focus on raw GPS speed data while some studies apply additional information, such as geographic information system (GIS) data, to obtain better detection performance. Procuring and maintaining dedicated road GIS data are costly and not trivial, whereas the technical maturity and broad use of map service application program interface (API) queries offers opportunities for mode detection tasks. The proposed driving cycle detection method takes advantage of map service APIs to obtain high-quality car-mode API route information and uses a trajectory segmentation algorithm to find the best-matched API route. The car-mode API route data combined with the actual route information, including the actual mode information, are used to train a logistic regression machine learning model, which estimates car modes and non-car modes with probability rates. The experimental results show promise for the proposed method’s ability to detect vehicle mode accurately.     

Adaptive traffic signal control with actor-critic methods in a real-world traffic network with different traffic disruption events

The transportation demand is rapidly growing in metropolises, resulting in chronic traffic congestions in dense downtown areas. Adaptive traffic signal control as the principle part of intelligent transportation systems has a primary role to effectively reduce traffic congestion by making a real-time adaptation in response to the changing traffic network dynamics. Reinforcement learning (RL) is an effective approach in machine learning that has been applied for designing adaptive traffic signal controllers. One of the most efficient and robust type of RL algorithms are continuous state actor-critic algorithms that have the advantage of fast learning and the ability to generalize to new and unseen traffic conditions. These algorithms are utilized in this paper to design adaptive traffic signal controllers called actor-critic adaptive traffic signal controllers (A-CATs controllers).The contribution of the present work rests on the integration of three threads: (a) showing performance comparisons of both discrete and continuous A-CATs controllers in a traffic network with recurring congestion (24-h traffic demand) in the upper downtown core of Tehran city, (b) analyzing the effects of different traffic disruptions including opportunistic pedestrians crossing, parking lane, non-recurring congestion, and different levels of sensor noise on the performance of A-CATS controllers, and (c) comparing the performance of different function approximators (tile coding and radial basis function) on the learning of A-CATs controllers. To this end, first an agent-based traffic simulation of the study area is carried out. Then six different scenarios are conducted to find the best A-CATs controller that is robust enough against different traffic disruptions. We observe that the A-CATs controller based on radial basis function networks (RBF (5)) outperforms others. This controller is benchmarked against controllers of discrete state Q-learning, Bayesian Q-learning, fixed time and actuated controllers; and the results reveal that it consistently outperforms them.     

A tensor-based Bayesian probabilistic model for citywide personalized travel time estimation

Urban travel time information is of great importance for many levels of traffic management and operation. This paper develops a tensor-based Bayesian probabilistic model for citywide and personalized travel time estimation, using the large-scale and sparse GPS trajectories generated by taxicabs. Combined with the knowledge learned from historical trajectories, travel times of different drivers on all road segments in some time slots are modeled with a 3-order tensor. This tensor-based modeling approach incorporates both the spatial correlation between different road segments and the person-specific variation between different drivers, as well as the coarse-grain temporal correlation between recent and historical traffic conditions and the fine-grain temporal correlation between different time slots. To account for the variability caused by the intrinsic uncertainties in urban road network, each travel time entry in the built tensor is treated as a variable following a log-normal distribution. With the help of the fully Bayesian treatment, the model achieves automatic hyper-parameter tuning and model complexity controlling, and therefore the problem of over-fitting is prevented even when the used data is large-scale and sparse. The proposed model is applied to a real case study on the citywide road network of Beijing, China, using the large-scale and sparse GPS trajectories collected from over 32,670 taxicabs for a period of two months. Empirical results of extensive experiments demonstrate that the proposed model provides an effective and robust approach for urban travel time estimation and outperforms the considered competing methods.     

A mixed Bayesian network for two-dimensional decision modeling of departure time and mode choice

The modeling of travel decision making has been a popular topic in transportation planning. Previous studies focused on random-utility discrete choice models and machine learning methods. This paper proposes a new modeling approach that utilizes a mixed Bayesian network (BN) for travel decision inference. The authors use a predetermined BN structure and calculate priori and posterior probability distributions of the decision alternatives based on the observed explanatory variables. As a “utility-free” decision inference method, the BN model releases the linear structure in the utility function but assumes the traffic level of service variables follow multivariate Gaussian distribution conditional on the choice variable. A real-world case study is conducted by using the regional travel survey data for a two-dimensional decision modeling of both departure time choice and travel mode choice. The results indicate that a two-dimensional mixed BN provides better accuracy than decision tree models and nested logit models. In addition, one can derive continuous elasticity with respect to each continuous explanatory variable for sensitivity analysis. This new approach addresses a research gap in probabilistic travel decision making modeling as well as two-dimensional travel decision modeling.     

Review of GPS Travel Survey and GPS Data-Processing Methods

Abstract

Global positioning system (GPS) devices have been utilised in travel surveys since the late 1990s. Because GPS devices are very accurate at recording time and positional characteristics of travel, they can correct the trip-misreporting issue resulting from self-reports of travel and improve the accuracy of travel data. Although the initial idea of using GPS surveys in transport data collection was just to replace paper-based travel diaries, GPS surveys currently are being applied in a number of transport fields. Several general reviews have been done about GPS surveys in the literature review sections in some papers, but a detailed systematic review from GPS data collection to the whole procedure of GPS data processing has not been undertaken. This paper comprehensively reviews the development of GPS surveys and their applications, and GPS data processing. Different from most reviews in GPS research, this paper provides a detailed and systematic comparison between different methods from trip identification to mode and purpose detection, introduces the methods that researchers and planners are currently using, and discusses the pros and cons of those methods. Based on this review, researchers can choose appropriate methods and endeavour to improve them.     

Integrity of map-matching algorithms

Map-matching algorithms are used to integrate positioning data with digital road network data so that vehicles can be placed on a road map. However, due to error associated with both positioning and map data, there can be a high degree of uncertainty associated with the map-matched locations. A quality indicator representing the level of confidence (integrity) in map-matched locations is essential for some Intelligent Transport System applications and could provide a warning to the user and provide a means of fast recovery from a failure. The objective of this paper is to determine an empirical method to derive the integrity of a map-matched location for three previously developed algorithms. This is achieved by formulating a metric based on various error sources associated with the positioning data and the map data. The metric ranges from 0 to 100 where 0 indicates a very high level of uncertainty in the map-matched location and 100 indicates a very low level of uncertainty. The integrity method is then tested for the three map-matching algorithms in the cases when the positioning data is from either a stand-alone global positioning system (GPS) or GPS integrated with deduced reckoning (DR) and for map data from three different scales (1:1250, 1:2500, and 1:50 000). The results suggest that the performance of the integrity method depends on the type of map-matching algorithm and the quality of the digital map data. A valid integrity warning is achieved 98.2% of the time in the case of the fuzzy logic map-matching algorithm with positioning data come from integrated GPS/DR and a digital map data with a scale of 1:2500.     

An analysis of the factors influencing differences in survey-reported and GPS-recorded trips

Recent advances in global positioning systems (GPS) technology have resulted in a transition in household travel survey methods to test the use of GPS units to record travel details, followed by the application of an algorithm to both identify trips and impute trip purpose, typically supplemented with some level of respondent confirmation via prompted-recall surveys. As the research community evaluates this new approach to potentially replace the traditional survey-reported collection method, it is important to consider how well the GPS-recorded and algorithm-imputed details capture trip details and whether the traditional survey-reported collection method may be preferred with regards to some types of travel. This paper considers two measures of travel intensity (survey-reported and GPS-recorded) for two trip purposes (work and non-work) as dependent variables in a joint ordered response model. The empirical analysis uses a sample from the full-study of the 2009 Indianapolis regional household travel survey. Individuals in this sample provided diary details about their travel survey day as well as carried wearable GPS units for the same 24-h period. The empirical results provide important insights regarding differences in measures of travel intensities related to the two different data collection modes (diary and GPS). The results suggest that more research is needed in the development of workplace identification algorithms, that GPS should continue to be used alongside rather than in lieu of the traditional diary approach, and that assignment of individuals to the GPS or diary survey approach should consider demographics and other characteristics.     

Repeatability & reproducibility: Implications of using GPS data for freight activity chains

As transport modellers we are interested in capturing the behaviour of freight vehicles that includes the locations at which vehicles perform their activities, the duration of activities, how often these locations are visited, and the sequence in which they are visited. With disaggregated freight behaviour data being scarce, transport modellers have identified vehicle tracking and fleet management companies as ideal third party sources for GPS travel data. GPS data does not provide us with behavioural information, but allows us to infer and extract behavioural knowledge using a variety of processing techniques. Many researchers remain sceptical as specific human intervention, referred to as ‘expert knowledge’, is often required during the processing phase: each GPS data set has unique characteristics and requires unique processing techniques and validation to extract the necessary behavioural information. Although much of the GPS data processing is automated through algorithms, human scrutiny is required to decide what algorithmic parameters as considered ‘best’, or at least ‘good’. In this paper we investigate the repeatability and reproducibility (R&R) of a method that entails variable human intervention in processing GPS data. More specifically, the judgement made by an observer with domain expertise on what clustering parameters applied to GPS data best identify the facilities where commercial vehicles perform their activities. By studying repeatability we want to answer the question ‘if the same expert analyses the GPS data more than once, how similar are the outcomes?’, and with reproducibility we want to answer the question ‘if different experts analyse the same GPS data, how similar are their outcomes?’ We follow two approaches to quantify the R&R and conclude in both cases that the measurement system is accurate. The use of GPS data and the associated expert judgements can hence be applied with confidence in freight transport models.     

Machine learning applications in activity-travel behaviour research: a review

ABSTRACT

This paper reviews the activity-travel behaviour literature that employs Machine Learning (ML) techniques for empirical analysis and modelling. Machine Learning algorithms, which attempt to build intelligence utilizing the availability of large amounts of data, have emerged as powerful tools in the fields of pattern recognition and big data analysis. These techniques have been applied in activity-travel behaviour studies since the early ’90s when Artificial Neural Networks (ANN) were employed to model mode choice decisions. AMOS, an activity-based modelling system developed in the mid-’90s, has ANN at its core to model and predict individual responses to travel demand management measures. In the dawn of 2000, ALBATROSS, a comprehensive activity-based travel demand modelling system, was proposed by Arentze and Timmermans using Decision Trees. Since then researchers have been exploring ML techniques like Support Vector Machines (SVM), Decision Trees (DT), Neural Networks (NN), Bayes Classifiers, and more recently, Ensemble Learners to model and predict activity-travel behaviour. A large number of publications over the years and an upward trend in the number of published articles over time indicate that Machine Learning is a promising tool for activity-travel behaviour analysis and prediction. This article, first of its kind in the literature, reviews these studies and explores the trends in activity-travel behaviour research that apply ML techniques. The review finds that mode choice decisions have received wide attention in the literature on ML applications. It was observed that most of the studies identify the lack of interpretability as a serious shortcoming in ML techniques. However, very few studies have attempted to improve the interpretability of the models. Further, some studies report the importance of feature engineering in ML-based studies, but very few studies adopt feature engineering before model development. Spatiotemporal transferability of models is another issue that has received minimal attention in the literature. In the end, the paper discusses possible directions for future research in the area of activity-travel behaviour modelling using ML techniques.     

Interactive geovisualization of activity-travel patterns using three-dimensional geographical information systems: a methodological exploration with a large data set

A major difficulty in the analysis of disaggregate activity-travel behavior in the past arises from the many interacting dimensions involved (e.g. location, timing, duration and sequencing of trips and activities). Often, the researcher is forced to decompose activity-travel patterns into their component dimensions and focus only on one or two dimensions at a time, or to treat them as a multidimensional whole using multivariate methods to derive generalized activity-travel patterns. This paper describes several GIS-based three-dimensional (3D) geovisualization methods for dealing with the spatial and temporal dimensions of human activity-travel patterns at the same time while avoiding the interpretative complexity of multivariate pattern generalization or recognition methods. These methods are operationalized using interactive 3D GIS techniques and a travel diary data set collected in the Portland (Oregon) metropolitan region. The study demonstrates several advantages in using these methods. First, significance of the temporal dimension and its interaction with the spatial dimension in structuring the daily space-time trajectories of individuals can be clearly revealed. Second, they are effective tools for the exploratory analysis of activity diary data that can lead to more focused analysis in later stages of a study. They can also help the formulation of more realistic computational or behavioral travel models.     

Micro-simulation of daily activity-travel patterns for travel demand forecasting

The development and initial validation results of a micro-simulator for the generation of daily activity-travel patterns are presented in this paper. The simulator assumes a sequential history and time-of-day dependent structure. Its components are developed based on a decomposition of a daily activity-travel pattern into components to which certain aspects of observed activity-travel behavior correspond, thus establishing a link between mathematical models and observational data. Each of the model components is relatively simple and is estimated using commonly adopted estimation methods and existing data sets. A computer code has been developed and daily travel patterns have been generated by Monte Carlo simulation. Study results show that individuals' daily travel patterns can be synthesized in a practical manner by micro-simulation. Results of validation analyses suggest that properly representing rigidities in daily schedules is important in simulating daily travel patterns. This revised version was published online in June 2006 with corrections to the Cover Date.     

Modeling time-of-day car use behavior: A Bayesian network approach

In this research, a Bayesian network (BN) approach is proposed to model the car use behavior of drivers by time of day and to analyze its relationship with driver and car characteristics. The proposed BN model can be categorized as a tree-augmented naive (TAN) Bayesian network. A latent class variable is included in this model to describe the unobserved heterogeneity of drivers. Both the structure and the parameters are learned from the dataset, which is extracted from GPS data collected in Toyota City, Japan. Based on inferences and evidence sensitivity analysis using the estimated TAN model, the effects of each single observed characteristic on car use measures are tested and found to be significant. The features of each category of the latent class are also analyzed. By testing the effect of each car use measure on every other measure, it is found that the correlations between car use measures are significant and should be considered in modeling car use behavior.