A Causal Approach for Modelling Spatial Dynamics * A Preliminary Report Mehul Bhatt Department of Computer Science La Trobe University Melbourne, Australia 3086. mbhatt@cs.latrobe.edu.au Abstract We propose a causal approach, involving events identified by their causes and effects, for the mod- elling of spatial dynamics. The suitability of situ- ation calculus as a high-level formalism for repre- senting and reasoning about spatial dynamics is ex- plored and the causal framework is formalised us- ing the same. A systematic illustration of the man- ner in which various aspects of axiomatic qualita- tive spatial calculi may be represented within the proposed causal framework is presented. The main advantage of this approach is that based on the structure and semantics of the calculus, computa- tional tasks such as planning, projection and expla- nation can be directly exploited. Within the spe- cialised spatial reasoning domain, these translate to spatial planning/re-configuration, simulation and causal explanation (i.e., inferring cause from obser- vations). Furthermore, given the qualitative nature of the spatial theory and the non-monotonic reason- ing capability within the formalism, the approach is also better suited at representing human-like abil- ities of common-sense reasoning with incomplete information. The main hypothesis underlying our approach is that an alternate causal perspective of existing qualitative spatial calculi using high-level tools such as the situation calculus is essential for their utilisation in diverse application domains such as intelligent systems, cognitive robotics and event- based and Temporal-GIS. 1 Motivation Most research in qualitative spatial reasoning has focussed on the development of spatial calculi that are representative of distinct spatial domains – mereotopology [Randell et al. 1992], orientation [Freksa 1992, Moratz et al. 2000], dis- tance [Hernandez et al. 1995], cardinal directions [Ligozat 1998] etc. Furthermore, there has been considerable progress toward efficient computational mechanisms for reasoning within the respective spatial domains (see [Cohn and Haz- arika 2001a] for a complete overview). However, relatively * Worshop on Spatial and Temporal Reasoning, IJCAI 2007. little work has explicitly addressed the need to develop or exploit existing representational apparatus that will facili- tate the use these spatial calculi in realistic application do- mains – alternate views of existing spatial reasoning tech- niques are essential if spatial calculi encompassing unique or integrated spatial domains are to be utilised in realis- tic application scenarios, e.g., as control mechanisms in ro- botic software/intelligent systems, representation of human- like spatial reasoning or decision-making abilities in real and/or simulated environments or even as explanatory mod- els within event-based and temporal-GIS (TGIS) applications where representing and reasoning about dynamic geospa- tial phenomena is of utmost importance. There are a wide range of formalisms that have been developed for represent- ing and reasoning about dynamically changing environments (e.g., situation calculus [McCarthy and Hayes 1969], events calculus [Kowalski and Sergot 1986]). The utility of such higher level representational formalisms (involving reasoning about actions and change) for the modelling of spatial dynam- ics cannot be taken granted – rather fundamental problems (e.g., Frame, Ramification, Qualification [Shanahan 1997]) relevant to modelling changing environments have been thor- oughly investigated in the context of the class of formalisms aforementioned. This has resulted in several non-monotonic extensions to classical symbolic approaches that are better suited for representing human-like abilities of common-sense reasoning with incomplete information. Furthermore, the is- sue of concurrent and continuous phenomena, which mani- fest themselves even in the simplest of dynamic domains, has been rigorously investigated in the context of the class of for- malisms developed within the area of reasoning about actions and change. This is especially useful for modelling of spa- tial dynamics, considering that the issue of concurrent spatial changes has not been addressed within the specialised spatial reasoning domain. Based on the situation calculus formalism, we propose a causal approach for the representation of spatial dynamics. The approach utilises an explicit notion of causality involving events that are identified by their causes and effects. Several distinctions of occurrences into internal and external events and actions are applicable. However, in this paper, only a lim- ited notion of an event based on primitive spatial transitions definable within in a qualitative spatial theory is used. For the spatial part of the theory, we operate within a purely region-