R4-[LogoSpace] - Details

Constraint-Based Reasoning in Qualitative Spatio-Temporal Calculi

 

The project R4-[LogoSpace] is concerned with constraint-based formalisms that allow for representing, and reasoning about, spatial and temporal information (QSTR). The aim of the project is to develop such formalisms, to analyze the computational properties of reasoning problems in these formalisms, and to develop and implement procedures to solve these problems.

In previous work the project developed methods for the combination of QSTR formalisms, studied qualitative reasoning in the context of multi-agent systems, and applied QSTR techniques to solve spatial planning tasks. For application and evaluation purposes, state-of-the-art reasoning techniques were implemented in a constraint solving library. Moreover, the project contributed to research on the problem of how humans process spatial information.

The current focus of the project is on enhancing QSTR methods in order to meet further requirements arising from application fields. A fundamental issue in this context is to integrate QSTR formalisms with other existing knowledge representation technologies and to develop methods that allows for using quantitative, metric information within qualitative reasoning methods. The project also investigates how recent advanced reasoning techniques from the finite CSP and SAT domains can be adapted to QSTR. Finally, we develop a unifying first order theory for QSTR as a glue between different constraint languages.

Our theoretical results are applied to solve exemplary reasoning tasks in the domains of spatial planning, human-machine interaction, and GIS. For example, the project studies spatio-temporal reasoning tasks that arise when geospatial data, such as historical data or traces from GPS devices, are queried or analyzed. In this context the project also considers default assumptions about spatial change in time.