Maximilian Leodolter

Personalized and Situation-Aware Multimodal Route Recommendations: The FAVOUR Algorithm

Route choice in multimodal networks shows a considerable variation between different individuals and the current situational context. Personalization and situation awareness of recommendation algorithms are already common in many areas, e.g., online retail. However, most online routing applications still provide shortest distance or shortest travel-time routes only, neglecting individual preferences, as well as the current situation. Both aspects are of particular importance in a multimodal setting as attractivity of some transportation modes, such as biking, which crucially depend on personal characteristics and exogenous factors, such as the weather. As an alternative, this paper introduces the FAVourite rOUte Recommendation (FAVOUR) approach to provide personalized, situation-aware route proposals based on three steps: first, at the initialization stage, the user provides limited information (home location, work place, mobility options, sociodemographics) used to select one out of a small number of initial profiles. Second, based on this information, a stated preference survey is designed in order to sharpen the profile. In this step, a mass preference prior (MPP) is used to encode the prior knowledge on preferences from the class identified in step one. Third, subsequently, the profile is continuously updated during usage of the routing services. The last two steps use Bayesian learning techniques in order to incorporate information from all contributing individuals. The FAVOUR approach is presented in detail and tested on a small number of survey participants. The experimental results on this real-world dataset show that FAVOUR generates better quality recommendations w.r.t. alternative learning algorithms from the literature. In particular, the definition of the MPP for initialization of step two is shown to provide better predictions than a number of alternatives from the literature.

Estimating travel times from static map attributes

Estimating accurate travel times on road networks is a prerequisite for many mobility related applications such as transportation planning, dynamic intermodal routing or logistics. Two widely used methods are (1) the trivial but inaccurate calculation of travel times using static speeds taken from road maps and (2) the use of historic time series calculated from rich data sets. For the second method extensive measurement campaigns are required. In this paper we present a novel approach to estimate realistic travel times exclusively from static map coefficients without the need for further data collection. Our method uses a linear regression model to estimate the diurnal variation of travel times for cars in urban and interurban areas. We discuss the model which has been developed and calibrated for the city of Vienna, Austria, and demonstrate the transferability of the model to a different city.

Improving vehicle speed prediction transferability with network centrality

Even though we are currently witnessing an unprecedented growth in the collection of movement data, practitioners in many fields still struggle with gaining access to reusable mobility data, such as traffic flows and speeds. Data availability varies considerably between different cities and regions. While some publish comprehensive open datasets, others either do not provide their data or do not even posses any traffic data. This paper proposes a solution to the problem of missing vehicle speed data. Our approach is to train a prediction model in an area where data is available and then transfer this model to areas where data is lacking. The proposed method requires only readily available static road network data in the target area. We improve upon previously published prediction models by incorporating local network centrality measures. Our approach reduces errors in vehicle speed prediction by as much as 24%.

Semi-supervised segmentation of accelerometer time series for transport mode classification

Collecting ground truth data with smart phone applications is as difficult as important for training classification models predicting transport modes of people. Errors of respondent input with respect to trip length and transport mode segmenting introduce a systematic bias in the classification model. We propose a semi-supervised framework adjusting user-given input to process user-collected accelerometer time series data. Our contributions are (1) an evaluation of the impact of segmentation bias, (2) a novel algorithm to find more homogeneous segments and (3) a robust incrementally trained classifier model based on clustering employing Dynamic Time Warping as similarity measure. We apply the proposed method on synthetic and real-world accelerometer trip data of 800 labeled trips consisting of 2000 user-given segments and 400 hours travel time and test it against a baseline classifier relying completely on user-feedback. The results prove that our method learns clusters revised from noise and increases the classifiers accuracy for real-world and synthetic data by up to 17%.

Personalization of Routing Services

Mobile technologies are improving constantly, and there are plenty of online routing services around. Although users are willing to try new services and apps, it is hard to retain them as users unless the service meets their expectations straightaway. In multimodal routing, using a generic model might not give appropriate routes to both an environmentally conscious traveler who uses mostly public transport and a car enthusiast. Following a trend in online retailing, in which personalized services are already the norm, research in the personalization of online routing services has shown promising results. However, because users tend to use a new service only a few times if the results are not satisfying right from the start, the initialization of the personalized model is of core importance. The authors propose a methodology that is based on a small initial data collection within the service and combined with two models that classify users into classes, first a latent class model and second a mixed logit approach with a prior clustering step. The authors tested the modeling approaches for their performance in the initialization stage and found that the classification approach can improve the correct detection of users’ chosen routes compared with state-of-the-art methodologies by several percentage points. Furthermore, user reactions to personalized routes included in a routing application were collected. The results show that the personalization of routes is well received by the users.

Improving vehicle speed estimates using street network centrality

ABSTRACT This paper describes a novel approach to improve prediction models which estimate vehicle speeds and their diurnal variation for road network links in urban street networks using only static map attributes. The presented approach takes into account previously neglected spatial information by integrating network centrality measures for closeness (indicating how central a link is) and betweenness (indicating how important a road link is) into the prediction model. The model is calibrated with a real-world dataset of 100 million individual speed measurements from a fleet of 3500 taxi probe vehicles in Vienna, Austria. Given that centrality can be derived directly from readily available street network data, the experimental results demonstrate that integrating centrality measures considerably improves the predictions without the need for adding a supplementary data source. Improvements for vehicle speed estimates are particularly prevalent on important street network links in the city center as well as rural streets in the periphery.

Influence of Weather on Transport Demand: Case Study from the Vienna, Austria, Region

In times of increasing travel demand, urban transport systems are under continuous stress. Knowledge of the impact of weather on any given day needs to be obtained for efficient operation of the system. Although it is expected that weather-related impacts will not dominate travel demand (e.g., work trips cannot be easily omitted), trips may be delayed or different modes may be chosen. It is well known that transport systems that operate close to capacity react highly nonlinearly to additional demand. Thus, changes in weather might lead to totally different settings for the management of the transport systems. This paper provides evidence of the influence of weather on travel demand for the greater Vienna, Austria, region. Long-term household mobility surveys were used for a descriptive analysis of the influence of weather on travel behavior. Statistical modeling of smaller mobility surveys allowed extrapolation to new situations, as well as an analysis of the joint influence of several variables. Significant evidence was provided that weather had a strong influence on the mobility choices of a large part of the population. The results emphasized that the weather impact depends heavily on the mode and purpose of the trip and the characteristics of the traveler. The results of the study can be used at the aggregate level to predict the impact of weather on traffic demand. This information is important for the management of transport system supply and demand, for ensuring the efficiency of urban transportation systems.