Maritime Context

The maritime environment still represents unexploited potential for modeling, management, and understanding of mobility data. The environment is diverse, open but partly ruled, and covers a large spectrum of ships, from small sailboats to supertankers, which generally exhibit type-related behaviors. Similarly to terrestrial or aerial domains, several real-time positioning systems, such as the Automatic Identification System (AIS), have been developed for keeping track of vessel movements. However, the huge amounts of data provided by these reporting systems are rarely used for knowledge discovery. This chapter aims at discussing different aspects of maritime mobilities understanding. This chapter enables readers to, first, understand the intrinsic behavior of maritime positioning systems and then proposes a methodology to illustrate the different steps leading to trajectory patterns for the understanding of outlier detection.

Maritime Traffic

The maritime environment has a huge impact on the world economy and our everyday lives. Beyond being a space where numerous marine species live, the sea is also a place where human activities (sailing, cruising, fishing, goods transportation, etc.) evolve and increase drastically. For example, world maritime trade of goods volume has doubled since the seventies and reached about 90% of global trade in terms of volume and 70% in terms of value. This ever increasing traffic leads to navigation difficulties and risks in coastal and crowded areas where numerous ships exhibit different movement objectives (sailing, fishing, etc.), which can be conflicting.

The Difficult Nature of Measuring Human Mobility

Human mobility on different spatial and temporal scales affects many processes taking place in our world. Although few will disagree that the large increase in human mobility during the twenty-first century has improved our general quality of life, it is increasingly confronting us with some of its more negative implications as well: congestion in and around densely populated areas by daily commuter traffic and the resulting strain on our environment, safety issues arising from the gathering of large crowds in relatively small areas, sudden risks of global pandemics and the difficulty of containing them, and so on. As such, an increase in human mobility should be accompanied by a deeper understanding of the processes governing these movements in order to better mitigate their negative implications.

A starting point in learning more about these movements is adequately measuring them. Until recently, this has been quite problematic. Qualitative methods such as shadowing and the collection of travel diaries are known to be error prone and labor intensive. An alternative method of tracking people in smaller-scale settings is through video surveillance systems. Despite technological advancements in the last decade, using cameras to reconstruct the movements of a large number of people in a realistic environment remains very difficult. Correctly identifying trajectories of individuals in one camera view is already nontrivial due to interactions between moving objects, changing illumination in outdoor environments, and so on.

Introduction

Mobility data are inherently uncertain due to several contributing factors related to different phases of their life cycle, from acquisition to interpretation. When data are processed, uncertainty propagates to intermediate and final results. Thus, it is important to be aware of uncertainty in trajectory data and explicitly account for it in their modeling and managing. For example, consider a simple scenario where people move around a city and disclose their positions twice an hour; to avoid stalking, the disclosed position is randomly selected from inside a circle with a radius of one kilometer, which contains the position of the user. Not being aware of uncertainty could lead to inconsistent conclusions. For instance, we could erroneously assume that a group of people have met or that someone has visited a privacy-sensitive place. On the contrary, taking uncertainty into account, we can avoid such erroneous conclusions; for example, if someone was farther than one kilometer from the place of an accident, we can certainly assume that this person was not involved in that accident.

We next introduce a well-known taxonomy of uncertainty (see Bibliographic Notes section), aimed at clearly defining terms that are often given multiple meanings in the literature.

A Taxonomy of Uncertainty

The taxonomy we present here considers, at the highest abstraction level, that uncertainty in mobility and geographic information is caused by the complexity of the system conformed by three kinds of entities: human being, earth (i.e., geographic/moving), and computing machinery.

Introduction

Visual analytics develops knowledge, methods, and technologies that exploit and combine the strengths of human and electronic data processing (Keim et al., 2008). Technically, visual analytics combines interactive visual techniques with algorithms for computational data analysis. The key role of the visual techniques is to enable and promote human understanding of the data and human reasoning about the data, which are necessary, in particular, for choosing appropriate computational methods and steering their work. Visual analytics approaches are applied to data and problems for which there are (yet) no purely automatic methods. By enabling human understanding, reasoning, and use of prior knowledge and experiences, visual analytics can help the analyst to find suitable methods for data analysis and problem solving, which, possibly, can later be fully or partly automated. In this way, visual analytics can drive the development and adaptation of computational analysis and learning algorithms.

Visualization is particularly essential for analyzing phenomena and processes unfolding in geographical space. Since the heterogeneity of the space and the variety of properties and relationships occurring in it cannot be adequately represented for fully automatic processing, exploration and analysis of geospatial data and the derivation of knowledge from it needs to rely upon the human analyst's sense of the space and place, tacit knowledge of their inherent properties and relationships, and space/place-related experiences. This applies, among others, to movement data.

Introduction

The research area of trajectory databases has addressed the need for representing movements of objects (i.e., trajectories) in databases in order to perform ad hoc querying and analysis on them. During the last decade, there has been a lot of research ranging from data models and query languages to implementation aspects, such as efficient indexing, query processing, and optimization techniques.

This chapter covers aspects related to data collection and handling so as to feed trajectory databases with appropriate data. We will also focus on the step trajectory reconstruction of the Geographic Privacy-aware KDD process (illustrated in Figure 2.1) emerged from the GeoPKDD project which proposed some solid theoretical foundations at an appropriate level of abstraction to deal with traces and trajectories of moving objects aiming at serving real world applications. This process consists of a set of techniques and methodologies that are applicable to mobility data and are organized in some well-defined and individual steps that have a clear target: to extract user-consumable forms of knowledge from large amounts of raw geographic data referenced in space and in time. However, when mobility data are about individuals, data collection is subject to privacy regulations and restrictions. To enable privacy-aware collection of position data, a complementary class of techniques is used, known as location PETs (privacy-enhancing technologies).

This KDD process can be applied to heterogeneous sources of mobility data.

Introduction

For a long time, applications have been using data about the positions of the moving objects they are interested in. For example, city planning applications, in particular in the transportation and traffic management domains, have been observing and monitoring traffic flows to capture their characteristics, namely their importance and localization, with the aim to build better models for traffic regulation and to identify solutions for future development of the existing road network. Sociologists have also been examining the movement of cars equipped with GPS, focusing on individual cars rather than traffic flows, to understand the habits of their drivers. In the logistics domain, applications have been monitoring the localization of the parcels during their transportation from their source locations to their destinations. These applications use the data both to be able to locate a parcel at any time and to optimize the performance of the transportation and distribution strategy. Similar concerns rule the management of data tracking airline passengers and their luggage. Ecologists have been observing animals and, whenever possible, tracking them via transmitters and satellites, mainly to understand animals' individual and group behaviors. Nowadays many enterprises are looking to extract information about their potential consumers out of the tracks left by their smartphones, electronic tablets, or access to social networks such as Flickr and Foursquare that record the geographic position of their users.

Traditionally, data about movement have been captured using static facilities, for example, sensors producing traffic flow measures or detecting an animal's presence. Data acquisition facilities changed drastically with the availability of embedded positioning devices (e.g., GPS).

Introduction

In previous chapters we have seen that the usage of location-aware devices enables the collection of large volumes of trajectory data. Effective analysis of such data imposes new challenges for their management, while raising opportunities for discovering behavioral patterns that can be exploited in applications such as location-based services or traffic control management.

Data warehouses (DW) and online analytical processing (OLAP) have been successfully used for transforming detailed data into valuable knowledge for decision-making purposes. Extending DWs for coping with trajectory data, leading to trajectory data warehouses (TDW), allows us to extract essential knowledge from raw or semantic trajectories. For example, a TDW can be used for analyzing the average speed of cars in different urban areas.

Trajectory data in a warehouse must be typically analyzed in conjunction with other data, for example, to find out the correlation between the speed of cars and temperature, precipitation, or elevation. In light of these needs, in this chapter we provide an overall view that integrates trajectory data in a more general data warehousing framework, which we call spatio-temporal data warehousing.

We start this chapter by introducing in Section 4.2 the notion of data warehousing and describing the main elements in a DW architecture. After giving in Section 4.3 the running example used throughout this chapter, we address in Section 4.4 spatio-temporal data warehousing, and show that trajectory data warehouses can be regarded as a particular case of spatio-temporal DW.

Introduction

The goal of air traffic control (ATC) is to maximize both safety and capacity, so as to accept all flights without compromising the life of the passengers or creating delays. Because air traffic is expected to double by 2030, new visualizations and analysis tools have to be developed to maintain and further improve the safety level. To do so, air traffic practitioners analyze data from the ATC activity. These multidimensional data include aircraft trajectories (3D location plus time), flight routes (ordered sequences of spatio-temporal points that represent planned routes), and meteorological data. In this chapter, we detail the relevant tasks of ATC practitioners and demonstrate recent visualization and query methods to fulfill them.

The special properties of ATC data propose new challenges and, at the same time, new opportunities of data analysis. The semantics of the data are rich because they includes the third dimension (altitude), which can be used to discover salient events such as takeoffs and landings. More semantics can be added by augmenting background data such as the traffic network and the meteorological data. ATC data sets are characterized by their large sizes, adding more challenges to the analysis. Trajectory analysis is difficult due to the data set size and to the fact that it contains many errors and uncertainties. One day's traffic over France contains about 20,000 trajectories (> 1 million records). Recording is done in a periodic manner (in our database: a radar plot, per aircraft, every 4 minutes), but a plot can be missed, or have erroneous data because of physical problems that occur at the time of recording.

From the invention of the wheel to moon-landing rockets, technological progress over thousands of years has produced increasingly powerful and efficient transportation means, thus making moving easier and easier. Most recent progress in telecommunications has added new facets to mobility. We have now the ability to automatically keep track of our travel routes and even document them with information such as photos about the places we have been. This prompted the surge of small to huge databases holding mobility data, that is, the data about where and when we have been all over the world as well as during our daily trips to reach our workplace. Complementarily, more and more applications in a great variety of domains have been or are being developed to make intelligent use of mobility data.

While most of us are aware that our cellphones and cars equipped with a GPS facility do regularly generate signals conveying their geographical position (plus other data characterizing movement, e.g., acceleration and instant speed), not everybody is aware of what may happen later to this data, that is, how it can be used, by whom, and for what purpose. This book aims to introduce the potential answers to this question. The presentation of the material aims to make the book an easy read for all professionals (students included) in computer sciences and geoinformatics. Special attention has been given to show enough examples to optimize the understanding of the discussions.

Introduction

Historical Overview

The curiosity of humans about animal movements dates back to ancient times and probably to prehistory. As a matter of fact, Aristotle (in. The History of Animals) described animal migrations. The capacity of animals to move with accuracy during long displacements was surprising and has been considered a mystery of nature till recent times. Much before the scientific foundation of diffusion due to the botanist Robert Brown in 1927, the roman poet Lucretius described in detail the motion of dust. For centuries, scholars hold Descartes' view that animals are thoughtless automata. Modern experimental research dates back to the end of nineteenth century, after the publication of The Origin of Species by Darwin in 1859. Researchers of that period adopted a subjective and anthropomorphic view of animal behavior and movements. Later, scholars started to interpret animal movement in a more objective, scientifically sound way, by investigating animal reactions to stimuli present in their environment, such as the gravitational field, the presence of light, gradient of humisdity, and so forth.

The concept that individual animals restrict their movements to finite areas known as home ranges is perhaps as old as ecology itself. Seton in 1909 observed that “No wild animal roams at random over the country; each has a home-region, even if it has not an actual home.” The definition of home range from Burt, dating back to 1943, is probably one of the most long lasting and widely used in ecology: “that area traversed by the individual in its normal activities of food gathering, mating and caring for young.

Introduction

Mobility data represent an invaluable source of information that can be recorded thanks to mobile telecommunications and ubiquitous computing where the locations of mobile users are continuously sensed. However, the collection, storage, and sharing of these movement data sets raise serious privacy concerns. In fact, position data may reveal the mobility behavior of the people: where they are going, where they live, where they work, their religion and so on. All this information refers to the private personal sphere of a person and therefore the analysis of mobility data may potentially reveal many facets of his or her private life. As a consequence, these kinds of data have to be considered personal information to be protected against undesirable and unlawful disclosure.

In the specific case of mobility scenarios, there exist two major different contexts in which the location privacy problem has to be taken into consideration: online location-based services and offline data analysis context. In the first case, a user communicates to a service provider his or her location to receive on-the-fly a specific service. An example of LBS is find the closest point of interest (POI), where a POI could be a restaurant. Privacy issues in the context of online location-based services have been already addressed in Chapter 2. In the second case, large amounts of mobility data are collected and can be used for offline data mining analysis able to extract reliable knowledge useful to understand and manage intelligent transportation, urban planning, and sustainable mobility, as already highlighted in previous chapters.