Schlüsselwörter
(Englisch)
|
Moving objects; movement; tracking data; trajectory simulation; movement ecology; animal ecology; random trajectory generator
|
Forschungsprogramme
(Englisch)
|
COST-Action IC0903 - MOVE: Knowledge Discovery from Moving Objects
|
Kurzbeschreibung
(Englisch)
|
Comparative evaluation of the performance of diverse methods for the representation, spatiotemporal analysis, and visualization of movement trajectory data is one of the objectives of COST Action IC0903. For the development and evaluation of movement analysis tools, simulated or synthesized trajectories are an indispensable component, since only synthetic movement data allow for the full control required for empirical evaluation experiments. The CASIMO project aims to develop generic, customizable trajectory generators that allow for the simulation of individual trajectories embedded in some application-dependent geospatial context. Context will be modeled as object-environment (o-e) relations and as object-object (o-o) relations (the latter will include the notion of .interactional' patterns). Besides developing geocontextaware trajectory simulation methods, the project will also aim to contribute sound evaluation methodologies that can be used in benchmark experiments of trajectory mining algorithms, as well as a set of diagnostic tools that can support evaluation. Finally, extensive experiments will help comparing the applicability of various trajectory simulation approaches empirically. The trajectory generators developed in CASIMO will differ from existing movement simulation procedures in three ways: They will be context-aware; they will be more generic and thus applicable to different types of moving objects; and they will allow accurately defining particular test movement patterns as benchmarks for pattern recognition algorithms.
|
Partner und Internationale Organisationen
(Englisch)
|
AT, BE, CH, DE, DK, EE, ES, FI, FR, GR, IE, IT, NL, PL, PT, RO, RS, TR, UK
|
Abstract
(Englisch)
|
The simulation of movement trajectories is one of the objectives of COST Action IC0903, serving purposes such as gap filling in incomplete trajectories, or the simulation of moving agents such as pedestrians moving in a city, or paths of long-haul movement of migrating birds. Hence, project CASIMO was launched in 2011, funding one PhD student, with the aim of developing a generic, customizable trajectory generator for the simulation of trajectories embedded in an application-dependent geospatial context. The focus of the third project year was on the development of an algorithm that can efficiently and reliably generate large numbers of trajectories between two given endpoints. The trajectories follow the random walk model and can thus be used as a null model in animal ecology applications, such as in the study of bird migration. They can be conditioned by control parameters, including available time budget or speed pattern. In order to support the simulation of long-haul movement, the algorithm takes into account the spherical shape of the Earth. The PhD student started his third year with a visit to the Max Planck Institute for Ornithology (MPIO), who collaborated on the algorithm development, helping with the requirements specification and with the provision of tracking data used to calibrate the control parameters of the algorithm. A first version of the algorithm has been implemented and tested. This initial version can be conditioned by the speed pattern extracted from historical tracking data (i.e. median or maximum speed) and already represents an advancement over existing random walk or Brownian bridge models. A paper describing the algorithm has been submitted to the Int. J. of Geogr. Inf. Sc., and a further paper with more results has been submitted to the conference GIScience 2014. The source code of the algorithm will be made available as open source. In the next step, the algorithm is extended so it can be conditioned by empirical probability distributions of ecologically relevant movement parameters, such as speed and turning angle, as well as their interactions, such that the movement and navigation capacities of a given species can be realistically replicated in the simulation. The design of the extended version of the algorithm is complete, and much of the implementation has been accomplished already to date. The extended version of the algorithm will allow supporting further studies. The first one will focus on data enrichment of coarsely sampled trajectories. Using fine-grained trajectories about a particular species as training data to parameterize the algorithm, detail of fine granularity can be filled in between the original coarse sample points. The second study will deal with obstacle avoidance, in order to test the performance of the algorithm against adding spatial constraints in the trajectory generation.
|
Datenbankreferenzen
(Englisch)
|
Swiss Database: COST-DB of the State Secretariat for Education and Research Hallwylstrasse 4 CH-3003 Berne, Switzerland Tel. +41 31 322 74 82 Swiss Project-Number: C09.0167
|
