Article Preview
Top1. Introduction
Building Information Modelling is most probably set to become the paradigm that will dominate the world of 3D building modelling in the near future. It is increasingly adopted in the construction industry due to several of its advantages. The most prominent one is probably the solution it provides to the interoperability issue between the numerous actors involved in a building construction process. Indeed, almost all of them need to rely on an abstraction of the building at some point for different purposes (visualization, planning, simulation, quantity takeoff, etc.). Due to their different points of view, the same building can be represented multiple times and differently, leading to laborious and error prone collaborations. The BIM paradigm makes it possible for the actors of the construction process to collaborate easier, by gathering all the needs of all the experts in a single model (Isikdag et al., 2012).
The most renowned open BIM standard is the Industry Foundation Classes (IFC) released by the International Alliance for Interoperability (IAI), now known as BuildingSMART Alliance (BuildingSMART, 2015). The IFC standard provides an elaborated description allowing to semantically and geometrically describe building components, as well as their spatial relationships. A huge amount of building components is considered through approximately a thousand of classes. IFC is adopted by the biggest architectural design software products, followed by more and more applications that rely on it to extract specific information and perform specific tasks.
Geometry, semantics and topology are the main information necessary to pretend to a reliable and flexible 3D indoor navigation system. By gathering all of them, IFC models are of big interest for that application domain. Thanks to the description of the indoor space partitions (known as the IfcSpace class), the openings and the spatial links between them, it is possible to create a basic connectivity graph, in a straightforward manner, which could support path computation. Figure 1 provides a good illustration of an IFC model and its original space units. Simply by taking the spaces as nodes and considering the intermediate elements linking the spaces (windows, doors, stairs, etc.) as the edges of the graph, a network covering the connectivity of the indoor space can be made. This concept is widely accepted by many applications and adopted by the OGC standard IndoorGML (Lee et al., 2014).
Figure 1. Example of an IFC model of a house (left) and its IfcSpace objects (right). The openings and stairs which link the spaces are also illustrated.
However, similarly to all BIM based applications, if indoor navigation relies on the information stored in an IFC model, the correctness of its operations will depend heavily on the validity of the provided information as well. Therefore, valid geometry, topology and semantic in the IFC models is fundamental. Since IFC is mainly a semantic oriented format, the focus made on the latter makes it less exposed to error, compared to geometry and topology. In order to limit that problem, the IFC standard includes rules for the description and the implementation of its classes. For example, the spatial information definitions rely on the ISO 10303-42, which corresponds to the geometric and topological representation definitions of the STEP standard (Pratt, 2001).
Unfortunately, practices show that rules of the standards might not be respected and validity issues on the described objects hamper the optimal use of BIM models. 3D models for indoor navigation need the indoor spaces to be represented as closed volumes with planar surface boundaries made of consistently oriented faces. But it is common to have open volumes, inconsistent orientation of the object faces or intersecting volumes. On the other hand, since they are not mandatory, the IfcSpace objects may not even be represented, in some models. All those issues represent a serious limitation to IFC-based indoor navigation.