Contents lists available at ScienceDirect Automation in Construction journal homepage: www.elsevier.com/locate/autcon BIMSL: A generic approach to the integration of building information models with real-time sensor data Miguel Alves a , Paulo Carreira a , António Aguiar Costa b,* a INESC-ID, Instituto Superior Técnico, Universidade de Lisboa, Portugal b CERIS, Instituto Superior Técnico, Universidade de Lisboa, Portugal ARTICLE INFO Keywords: Building information modeling BIM sensor language Internet of things Sensors Real-time data Domain-specific language ABSTRACT The surge of interest in digital building models combined with increasing sensorization of spaces is prompting an integration of Building Information Modelling (BIM) with real-time sensor data. However, current approaches reported in literature either remain theoretical or target very specific application domains, showing that there is no generic approach to assist in the creation of service and application software that combines sensor data with BIM models. The solution presented in this article addresses the engineering complexity associated with integrating sensor data with BIM by leveraging an advanced software engineering technique known as a Domain-Specific Language (DSL). We demonstrate also that the language herein proposed, named Building Information Modeling Sensor Language(BIMSL) provides substantial gains of expressiveness and ease of use in developing queries that process sensor data with complex conditions over BIM models. BIMSL is validated with experienced software developers according to a consistent evaluation protocol for DSLs focused on effectiveness, efficiency, satisfaction, and usability attributes. The results outperform the standard existing alternatives, indicating that our proposal contributes to reducing the human effort associated with integrating BIM with sensor data. 1. Introduction The evolution of the Internet of Things is leading to an increase in sensorization of physical spaces [1], resulting in a growing number of applications that require reasoning about their surrounding environ- ment in real-time [2]. These applications benefit greatly from integra- tion with BIM models [1], which makes it possible to digitally represent physical and functional characteristics of physical spaces [3–5] and may, thereby, provide relevant information about the buildings. The above-mentioned integration simplifies the engineering process underlying the combination of sensor data with the physical elements and constructive characteristics of the building (required, for instance, to compare the measured performance with the expected performance of the designed building). Conceivably, an integrated approach to handle BIM models and sensor data will bring about a more structured development practice as the engineer conceiving the system has a more precise understanding of the entire data framework. Despite being highly regarded by the scientific community [6], BIM does not yet have strong capabilities to process real-time data [5], as do other solutions for sensors in smart environments. A number of chal- lenges have to be overcome in order to integrate real-time data with BIM models. One of the main challenges is to retrieve specific in- formation out of BIM [7,8], especially by using the Industry Foundation Class (IFC) product model [9], which is the most widely recognized model format to capture information of BIM models [8]. Another challenge has to do with the fact that real-time sensor data must be continuously processed, which involves complex processing techni- ques [10–12]. As we will make clear later (Section 2), overall, com- bining BIM and real-time data processing is a complex undertaking. Several approaches have been proposed to solve above-mentioned challenges of integrating BIM with real-time data [5,13-15]. However, most of them lack practical validation [14,16] or are highly dependent on domain specifications [17]. In fact, several approaches are related to the development of applications concerning specific application do- mains, namely (i) Energy Management (EM) [5,18-21],(ii) Building Automation (BA) [22–24],(iii) Fire Control [25],(iv) Health and Safety [15],(v) Safety Risk [26], and (vi) Augmented Reality [27]. None of these is adaptable to a wide range of application domains. The development of applications that combine BIM with real-time sensor data, similarly to other data-intensive applications, is hampered by the difficulty to specify correct queries that explore the data sources and feed data into the application. In our particular setting, the queries http://dx.doi.org/10.1016/j.autcon.2017.09.005 Received 7 November 2016; Received in revised form 3 September 2017; Accepted 7 September 2017 * Corresponding author. E-mail addresses: miguel.bastos.alves@tecnico.ulisboa.pt (M. Alves), paulo.carreira@tecnico.ulisboa.pt (P. Carreira), aguiar.costa@tecnico.ulisboa.pt (A.A. Costa). Automation in Construction 84 (2017) 304–314 Available online 28 September 2017 0926-5805/ © 2017 Elsevier B.V. All rights reserved. MARK