Exploring context in spatio-temporal data

点击翻译按钮获取中文摘要

Humans and animals often present movement behavior that can modeled as patterns. The long-term objective of this research is concerned with such movement patterns, i.e., trajectory patterns that can observed, grouped, summarized and indexed for further analysis. Each of those tasks pose interesting short- to medium-term research challenges. To give but a simple example application, assume that customary routes taken by a set of drivers are known and consider the event of an accident or road construction. Being able to determine in an efficient way which drivers would be affected and therefore proactively advising them of the potential impact of that events has a clear practical value. Note that checking every stored trajectory against each event of potential interest is clearly not practical. As well, handling only the spatial component of trajectories is not enough, e.g., if an accident happens in the morning it is very unlikely to affect those who drive through that location every evening. A different application domain where knowing trajectory patterns is useful is that of routing queries on-demand through mobile objects. For instance, in many application domains, such as farming, having the objects broadcast their data continuously is either too expensive (in terms of energy cost) or unfeasible (due to distances between objects and a central location). We envision a model of query dissemination, where objects are used as the transportation means to take queries from source to destination and back. In this case being able to take advantage of trajectory patterns is clearly an advantage in terms of query routing optimization. Finally, both application scenarios above become much more complex if probabilistic trajectories are considered instead of deterministic ones. We plan to start our research program by considering deterministic trajectories and move naturally towards probabilistic trajectories.