The first stage of this research aims at providing an overall understanding of movement itself and the patterns of movement, identifying the fundamental ele- ments of movement patterns. The substance of this stage is presented in detail in Research Paper 1 (see Part II, Research Paper 1, page 81).
ä Research Paper 1:
Dodge, S., Weibel, R. and Lautenschütz, A.K. (2008). Towards a Taxonomy of Movement Patterns. Journal of Information Visualization, Vol. 7, pp. 240 – 252.
3.1.1 Objectives
The main objectives of this stage are captured in Objective 1 and Objective 2 of this thesis:
Objective 1: This research shall develop a conceptual framework, encapsulating essential elements that characterize the movement behavior of objects. Objective 2: This research shall establish a comprehensive classification of move-
ment patterns. The identified movement patterns shall be defined employ- ing the elements of the conceptual framework.
Accordingly, Research Paper 1 aims at providing a framework identifying the fundamental elements of movement. Furthermore, it responds to the need for a catalog of movement patterns to facilitate the development of movement data mining techniques.
3.1.2 Methods and Results
We approached the objectives of this stage with a twofold process:
(a) bottom-up approach, by decomposing movement into its constituting ele- ments. The aim here was to develop a conceptual framework of the movement of dynamic objects that could be used to build definitions of individual move- ment patterns. The developed framework consists of the following elements: • Movement parameters are measurable quantities (i.e. primitives) and their derivatives such as position, time, speed, distance, direction. Move- ment parameters are categorized into spatial, temporal, spatio-temporal dimensions (Table 3.1).
• Number of moving objects involved (i.e. an individual object, a group of objects with a functional relationship, a cohort of objects with a common characteristic)
• Path type (i.e. continuous and discontinuous paths)
• Influencing factor such as intrinsic properties, spatial constraint, envi- ronment, and influence of other agents.
• Scale and granularity (i.e. spatial and temporal scales)
(b) top-down approach, with a survey of the research conducted prior to this study. The aim was to categorize patterns of movement proposed by other researchers and to discover commonalities and differences in the terminology and pattern types. The survey encompasses the available pertinent litera- ture in geographic knowledge discovery, data mining, and visual analytics on movement data, as well as additional application specific references.
The result of this step is the classification of movement patterns presented in Figure 3.1. At the top level of this classification, two classes of generic and behavioral patterns, respectively, were identified. Generic patterns refer to the low-level patterns that are usually insufficient to explain specific move- ment behavior of a particular object. Hence, the proposed classification of movement patterns distinguishes the higher-level behavioral patterns for the domain-specific applications. The generic patterns are further categorized into two classes of primitive patterns and compound patterns. Furthermore, the classification distinguishes three dimensions for generic movement pat- terns: spatial, temporal, and spatio-temporal patterns.
Table 3.1: Movement parameters
Dimensions Parameters
Primitives Primary derivatives Secondary derivatives Spatial Position (x, y) Distance f(posn) Spatialf (distance)distribution
Direction f(posn) Change of direction f (direction)
Spatial extent f(posn) Sinuosity f(distance) Temporal Instance (t)Interval (t) Duration f(t)Travel time f(t) Temporal distributionChange of duration
f (duration)
Spatio-temporal – Speed f(x, y, t) Acceleration f(speed)
Velocity f(x, y, t) Approaching rate
In order to facilitate the future development and consolidation of the proposed classification and pattern definitions into a complete taxonomy of movement pat- terns, a wiki1 was set up. The wiki hosts more background information and
more detailed definitions of movement patterns, and serves as an open discussion platform for the community.
3.1.3 Main Findings
• The occurrence of patterns can be influenced by various factors that impact or constrain the movement of objects. Examples of such factors include:
– intrinsic physical and behavioral properties of the moving object (e.g. saccadic movement of mass-less eyes in contrast to the very slow and smooth movement of a massive body such as an elephant)
– spatial constraints such as topography, road networks, and natural barriers (e.g. the occurrence of a moving cluster along a road). – environment (e.g. the evolution of hurricanes is influenced by the air
and sea temperatures and the air pressure (Elsner and Kara, 1999)) – influences of other agents can cause the emergence of certain pat-
terns in domain-specific applications (e.g. attraction and competition among animals lead to particular behavioral patterns such as courtship or fighting, respectively)
• Movement patterns can be classified into two main categories ; (a) generic patterns and (b) behavioral patterns (Figure 3.1). The generic patterns represent the building blocks used to form the behavioral patterns.
– Generic patterns can be found in any form of behavior that builds on movement of objects such as moving clusters, concurrence, and
Movement patterns
Primitive patterns Compound patterns Spatial Spatio-temporal Temporal Spatio-temporal
Spatial density Co-location in space Ordered Order- irrelevant Symmetrical Moving cluster Meet Fixed Varying Fixed Varying Incidents Concurrence Opposition Dispersion Co-incidence in space and time
Full Lagged Constancy Spatio-temporal sequence Spatio-temporal periodicity Synchronization Full Lagged Temporal sequence Temporal periodicity Temporal relations Trend/ fluctuation Propagation Convergence/ divergence Encounter/ breakup Trend-setting Symmetry
Generic patterns Behavioral patterns
Pursuit/ evasion Fighting Play Courtship Migration Parental protection Foraging Congestion Saccade/ Fixation Flock Loner Isolated MPO Repetition Hot spots
Figure 3.1: Classification of movement patterns
repetition patterns.
– Behavioral patterns correspond to a particular behavior typical of a particular object such as pursuit, evasion, or foraging behaviors of animals.
• Generic movement patterns exhibit different degrees of complexity: – Primitive patterns, which are the basic patterns that are formed as a
result of the similarity in the variations of a single movement param- eter, such as concurrency in speed or direction.
– Compound patterns, which are formed in the relations between multi- ple objects as a composition of several primitive patterns. For instance, the convergence of the movements of a set of objects to the same lo- cation, where the original movement direction of the involved object remains unchanged (Gudmundsson et al., 2007).
• Primitive patterns can occur purely in the spatial domain or in the tem- poral domain, or they can be mixed spatio-temporal. Behavioral patterns invariably involve both spatial and temporal dimensions.
3.1.4 Contributions
This work contributes to the development of a catalog for knowledge discovery and data mining algorithms by developing:
(a) a conceptual framework encapsulating the fundamental elements of the move- ment behavior of different objects; and
(b) a comprehensive classification and review of movement patterns (illustrated in Figure 3.1).