Ontic for autonomous drone flight guided by a digital construction modelAbstract
The present work aims to demonstrate the technical feasibility of a sustainable knowledge modeling process for building inspections. The following elements are integrated: knowledge bases for the Semantic Web; information contained in digital building models; the plotting of routes for autonomous drone flights, guided by spatial data from the model; and extensible ontologies that accumulate factual knowledge. Each autonomous flight route is plotted based on spatial data extracted from the digital model of the headquarters of an educational building, which contains semantic definitions of the building maintenance process. The routes are structured into WayPoints enriched with data from various semantics and saved in files compatible with georeferencing systems. The process proposes the creation of a factual knowledge base grounded in ontologies within the open and linked data paradigm. Initiatives such as the B-IRO model (Building-Input Relationship Output) are highlighted. The text presents a reflection on two intrinsic challenges of the proposal: the definition of ontologies and the need to build a network infrastructure with Semantic Web functionalities within the open and linked data paradigm. Both factors are necessary to sustain the process over time.
Keywords
Semantic modeling; UAV; UAS; Drone; KML; BIM; Ontologies
Resumo
O presente trabalho tem por objetivo demonstrar a viabilidade técnica de um processo sustentável de modelagem de conhecimento para vistorias prediais. São integrados os seguintes elementos: bases de conhecimento para a Web-semântica; informações contidas em modelos digitais da construção; o traçado de rotas para voos autônomos de drones, direcionadas pelos dados espaciais do modelo; e, ontologias extensíveis que acumulem conhecimento factual. Cada rota de voo autônomo é traçada com base nos dados espaciais extraídos do modelo digital da sede de um edifício educacional, que contém definições semânticas do processo de manutenção predial. Os percursos são estruturados em WayPoints atributados com dados provenientes de diversas semânticas e salvos em arquivos compatíveis com sistemas de georreferenciamento. O processo propõe a formação de uma base de conhecimento factual fundamentada em ontologias e no paradigma de dados abertos e vinculados. Se destacam iniciativas como a do modelo B-IRO (Building-Input Relationship Output). O texto apresenta uma reflexão sobre dois desafios inerentes à proposta: a definição de ontologias e a necessidade de construir uma infraestrutura de rede com funcionalidades de web-semântica no paradigma de dados abertos e vinculados. Ambos os fatores necessários para sustentar o processo no tempo.
Palavras-chave
Modelagem semântica; UAV; UAS; Drone; KML; BIM; Ontologias
1 Introduction
Unmanned Aerial Vehicles (UAVs), or simply drones, have been used to perform building inspection services by infrastructure maintenance teams (Peinado et al., 2025). In Brazil, the National Civil Aviation Authority (ANAC), the Department of Airspace Control (DECEA, 2023) and ANATEL use the acronym UAS, which in English means “Unmanned Aircraft System”. The three organizations have been developing regulations and recommendations for the responsible use of these devices. The acronym UAS is used to refer to three components of this technology, identifying the aircraft, the control system and the communication link. With a camera installed on the drone, it is possible to capture imagery that enables the creation of a historical, temporally ordered record of a built asset’s condition, providing material for management analysis to support maintenance decisions throughout the building’s life cycle.
Inspections carried out through photographic records or videos captured by the drone onboard cameras simplify the inspection process of roofs, façades and external building areas, and allow photogrammetry techniques to generate point clouds using the photographic material collected during the process. Peinado et al. (2025) reports the use of planned flights executed manually during the construction phase. Although drone flights can be performed manually, the literature highlights several systematic approaches for flight optimization (Altinses et al., 2024; Pan et al., 2025). Some researchers have investigated building operation and maintenance processes using systematic approaches (Jiao et al., 2023).
In this context, the capture of large quantities of images disconnected from their originating elements becomes a factor that complicates the planning of systematic processes related to maintenance operations. Conversely, maintaining a link that relates images to their original objects supports the decision-making process. This article discusses the conceptual decisions made during the development of an application programming interface (API) designed to perform autonomous drone flights, using spatial information extracted from the Building Information Model (BIM) of the inspected asset as guiding material. To define the proposed technique, three aspects were considered:
-
algorithms for segmentation and semantic integration of the BIM model and the flight plan;
-
an asset management system using the Building-Input Relationship Output (B-IRO) model introduced by Jiao et al. (2023); and
-
a study of the relationship between the B-IRO model and the formulation of ontologies formalized in RDF triples, aimed at integration within the Semantic Web as introduced by Berners-Lee (2009).
2 Theoretical assumptions
This work assumes that the building’s primary geometry remains stable throughout its life cycle. Under this premise, it becomes reasonable to consider that pre-programmed drone flights can be catalogued and reused to perform maintenance operations consistently over the asset’s entire life cycle, eliminating the positional variations typically introduced by manually piloted missions. In practical terms, once a flight plan is designed, calibrated, and validated, no further adjustments will be required. Each new inspection contributes fresh imagery to the historical record of past maintenance activities. Over time, the material collected in successive flights progressively forms a relational repository of information that is continuously enriched as data accumulates.
A second assumption is that the images generated during inspections should remain intrinsically linked to the objects that produced them. For this reason, the systematic planning of drone flights through controlled route formulation finds in the digital model an appropriate and reliable source of spatiotemporal information. In this approach, the building’s model becomes the driver of the flight plan: it supplies the spatial reference markers used to define the routes, as well as the set of non-geometric parameters (dates, names, codes, etc.) associated with the building across its design, construction, and operational phases. These parameters, which semantically characterize the asset, support the systematization of the temporal and logical aspects of operation and maintenance processes, ensuring that inspections remain coherent, repeatable, and contextually meaningful.
The application was developed in C# within the .NET environment to run on Revit 2024. It enables planning the geometric layout for flight routes by leveraging spatial data and semantic information embedded in the BIM model for the Technology Center at the Federal University of Rio de Janeiro (CT-UFRJ). The building’s morphology, established during its design and construction phases and finalized in 1965 (Santos, 2024), shapes the trajectories of drone flights. In other words, aerial circulation paths are conditioned by the architectural and structural logic of the design. The implemented drone routing geometries included linear trajectories, perimetral paths, regular matrices, irregular trajectories, sequential zigzags, and panoramic routes. The parameters associated with the model objects are used to qualify the flight (room names, window codes, column numbering, etc.). In the CT-UFRJ model, each construction component is classified according to pre-established institutional management codes and is planned by technical design and maintenance teams.
The spatial position of each construction instance in the BIM model is used to define two location points: the drone’s positioning point and the focal point for the drone camera’s line of sight. In façade inspection flights, for example, the BIM model data associated with each window helps define the drone’s positioning point (in front of the window) and the camera’s focal target (directed toward the window). In panoramic routes with circular or elliptical trajectories, the geometric position of a column can serve as the center of the route, while its non-geometric attributes provide the necessary control metadata. Therefore, each constructive element contributes to both aspects, spatial position and semantic information, enabling the automatic generation of precise, repeatable, and context-aware flight paths. For this purpose, each spatial coordinate is converted from its Cartesian representation into a geographic coordinate in the WGS84 Geodetic Reference System. This conversion defines the WayPoints of the planned route as well as the rotational orientation required for the camera during flight. To perform the transformation from Cartesian coordinates to geographic coordinates (longitude and latitude) was employed the .NET CoordinateSharp library.
The data required to manage each constructive or functional component — such as windows, columns, rooms, or sectors — are extracted, reorganized, and recorded as extended metadata associated with their corresponding WayPoints in a Keyhole Markup Language file (KML), compatible with geospatial systems. This ensures that each point along the route carries not only spatial information but also the semantic attributes, relevant to the inspection process. Drone technology encompasses a wide variety of device models and proprietary ecosystems. To ensure compatibility, the application was designed to structure the data in a format readable by the DJI Pilot application for Android, enabling seamless integration between the generated KML files and the operational software used during flight execution.
3 Method
To map the building, several flight geometries were programmed with different types of movement. These predefined trajectories were designed to accommodate the various spatial configurations and inspection needs of the built asset. The following movement types have been planned so far:
-
linear route with the camera facing horizontally and perpendicular to the flight path;
-
irregular zigzag route with the camera facing downward;
-
panoramic circular or elliptical route with the camera facing outward;
-
panoramic circular or elliptical route with the camera facing inward;
-
matrix-pattern route with the camera facing downward; and
-
perimeter route with the camera facing inward.
Each movement typology corresponds to a specific inspection objective. For example, when mapping positions related to the building’s interior rooms, an irregular zigzag route can be employed to create a sequential path that passes over all rooms. This type of trajectory is particularly useful before slab closure during construction, or for detecting anomalous conditions on roofs by identifying the spaces located directly beneath the anomaly.
All flights begin at a designated base area on the campus, selected for being free of obstacles such as trees, buildings, or overhead power lines. The mission starts with a vertical ascent to a safe altitude, after which the drone travels to the first Point of Interest (POI). It then executes the planned route at a constant altitude and concludes by returning to the base point, where it performs a controlled descent. At the current stage of development, changes in altitude occur only during takeoff and landing. However, a more advanced routing strategy - which allows altitude variation at any point along the trajectory - is currently under study to support more complex inspection scenarios and to improve the coverage of three-dimensional spaces. Figure 1 shows four different flight typologies departing from the planned takeoff and landing position within the CT-UFRJ campus area and data of one Waypoint.
3.1 KML file organization
Keyhole Markup Language files allow geographic data to be recorded and visualized on digital maps. KML provides well-defined semantics for fields that represent geographic information map positions, as well as for open and closed polygons used to trace routes or delimit areas. In the proposed workflow, the flight path is written by the application, which extracts and interprets objects from the BIM model and arranges them into a sequential movement plan. Each flight position is then registered in the KML file, which incorporates three specific semantic layers that enrich the representation of the trajectory and its operational context: (Figure 2)
-
the semantics of the KML information schema;
-
the semantics related to the drone (in this example, DJI); and
-
the semantics defined in the BIM model of the inspected asset (Figure 3).
In this example, the tags <Placemark>, <name>, <visibility>, <Point>, <altitudeMode> and <coordinates> correspond to the KML semantics. All data enclosed in tags beginning with the mis: prefix correspond to the semantics defined by the DJI drone manufacturer. The information contained within the tag <description> (such as Bloco: A, Tipo: WC 2845416, etc.) consists of semantic values associated with the BIM model object that defined the position of the flight point or its institutional identification. Additional fields are included to record flight-related metadata, such as timestamps, or to specify stylistic attributes, including colors, point types, and line formats. These elements help create a graphical representation of the route, aiding in visualizing the camera’s behavior throughout the flight path.
3.2 Extensible ontologies in BIM
As explained in the previous section, this workflow involves handling multiple information structures, each with its own semantics. For example, both the DJI drone model and the KML format have clearly established semantic rules. It is also possible that an institutional BIM model does not yet have a formally defined semantic structure for its objects, even though it will inevitably inherit some level of semantics from the modeling tool’s data structure or from a neutral standard such as Industry Foundation Classes files (IFC). In such cases, the absence of an explicit institutional semantic layer requires additional effort to align BIM objects with the semantic requirements of the inspection workflow. In this work, an ontological strategy was adopted to address the diverse semantics involved. In the field of architectural and engineering design, it is common to use visual programming languages to complement and expand the functionalities of project applications. Two of the most widespread languages, DesignScript and Grasshopper, allow us to work within the paradigms of functional, imperative, and associative programming. They open the possibility of automating parameterized processes and are widely recognized for facilitating the creation of more flexible design workflows, enabling designers’ thinking to evolve during the conception process (Aish, 2013).
However, these languages are not prepared to express the factual knowledge of the designed object. This function can be assumed by other sorts of languages that allow ontological formalization, which expands the field of thought implicitly incorporated into the project. Even more so during the design phase, when it is possible to express both the ideal and the factual determinations conceived for the project and to do so in a logical and computational manner. Thus, ontologies written for the technical domains of construction are a means to continue enriching the semantics of projects (Al-Hakam; Scherer, 2020). In addition to broadening the range of meanings and higher-order relationships between construction objects and the implicit architectural morphology of the model, they also allow the incorporation of mechanisms to automate logical reasoning operations associated with applications that assist designers. Some extensible ontologies to complement and enrich processes within BIM technology have begun to be studied and developed with greater attention in recent years.
In this field, notable works include: BOT (Building Topology Ontology), which proposes an axiomatic framework for BIM (Rasmussen; Hvid; Karkshoj, 2017); BROT (Bridge Topology Ontology), based on BOT but oriented toward urban infrastructure projects; the Brick ontology, which proposes a semantic for sensors and IoT devices in buildings, aligned with BOT; SAREF (Smart REFerence Ontology) (Poveda-Villalón; Garcia-Castro, 2018); The buildingSMART ifcOWL initiative, which incorporates the missing ontological layer into IFC data schemas (Bonduel et al., 2018). A more recent project is BIM4Ren (BIM for Renovation) (BIM4REN, 2024), which conducts studies pointing to the formalization of complex informational architecture, aligning domain ontologies within the Semantic Web paradigm. Funded by the European Union, its purpose is to incorporate computational intelligence into building renovation processes (Schulz; Oraskary; Beetz, 2023; BIM4REN, 2024). Our study did not identify projects of this kind within the regional contexts of Brazil and Argentina; however, this does not imply that such initiatives are absent. A detailed Systematic Literature Review (SLR) could be undertaken to highlight updated domain ontology-based BIM systematization projects in the region. Nevertheless, such an analysis lies beyond the scope of the present work, which focused on the development and integration of the ontology constructor, the design of UAS flight operations integrated with the Revit BIM model semantics and the conceptual definition of UAS ontology concepts for maintenance activities.
Since the formalization of extensible ontologies originates from cognitive sciences and computer science, there is still some difficulty in engaging designers in the process of modeling knowledge bases oriented to BIM and supported by ontological axiomatics. Generally, designers receive ready-made information schemas and adapt to them. They know the design objects, the commands of CAD or BIM applications, and ways to automate processes, but they struggle to make explicit the knowledge accumulated during the various stages of the building life cycle, such as the conception phase, modeling phase, construction, or maintenance works. According to Bus et al. (2018), building an ontological system over a project is not a trivial process. The task requires continuous exercise of epistemological reflection to express the ideal and factual knowledge of things. Building an ontology means theorizing not only about how to design, but also about how to guide work tools (information systems) so they can process the implicit layers of knowledge used during the design act, incorporated into the digital model (Young et al., 2007). Therefore, epistemological and practical aspects are involved. One of the lessons learned by the authors of the SAREF project states that systematic and consistent extraction of data from a model is hindered if the design team lacks domain specialists (Poveda-Villalón; Garcia-Castro, 2018). To this observation, one may add that data can only be systematically extracted once they have been systematically inserted. In other words, building ontologies oriented towards the design conception phase could not only facilitate later information extraction operations but also help materialize an ontogenetic description of each unique project, which will accompany the building throughout its entire life cycle.
3.3 The first challenge: building ontologies
To address the problem of creating and managing ontologies, a methodology was developed and a software tool was implemented to facilitate the task of writing ontologies in OWL/TTL syntax (Menegotto, 2025). Although, for pragmatic reasons, the program has been classified as an ontological constructor, it is hypothesized that the term ‘ontology’ may not be the most suitable designation for the task of formalizing OWL files in the domains of architecture and engineering. It is important to note, however, that the explanation provided hereafter refers only to the domains of architecture and engineering, which are characterized as creative domains in which elements of ideation are present. Domains that observe and describe reality more directly, such as biology, whose ontologies are guided by an Aristotelian epistemology, according to Smith’s (2008) realism, do not constitute the object of this hypothesis. The work of architects and engineers establishes a practical limit within the domain of representing factuality, understood as successive layers of facts accumulated over a period that is inherently creative. The development of the ontology constructor follows the five classical recommendations proposed by Gruber (1995) for ontology formalization, listed as follows:
-
Clarity: conceptual definitions must be necessary and sufficient, avoiding ambiguities. The ontology should allow any reader to clearly understand the meaning of the terms;
-
Coherence: formal descriptions cannot contradict informal descriptions written in natural language. The ontology must maintain logical consistency among all its elements;
-
Extensibility: new concepts must be added monotonically, that is, without breaking or altering the existing structure. More specific concepts should be contained within more general sets;
-
Minimal encoding: ontology should use as little implementation bias as possible, preferring simple syntaxes that are close to natural language, avoiding excessive use of symbols to improve readability; and
-
Minimal ontological commitment: objects should be conceptualized in the simplest possible way, avoiding unnecessary commitments about the nature of entities. This reduces bias and facilitates ontology reuse.
Complementing Gruber’s (1995) recommendations with Smith’s (2008) realism and Bryant’s (2017) philosophical investigations, the proposed hypothesis maintains that the representational task carried out in this work could be defined and designated as the implementation of an ontic BIM constructor. Thus, OWL files are understood not only as the Ontology Web Language, but also as the Ontic Web Language. In the following sections, this hypothesis will be examined in greater depth, as it is central to understanding the technical challenges involved. The technique seeks to provide a means of representing knowledge throughout the process of architectural becoming, in other words, to sustain this representation challenge throughout its entire life cycle.
3.4 The construction of an ontology for drone flights
As noted earlier, the application presented depends heavily on the semantics defined during the BIM modeling phases of the asset inspection. It is not unusual for semantic definitions to be missing in a building’s BIM model. This situation is understandable, since some necessary movements—such as changing altitude along the route to inspect adjacent roofs—are not evident during the initial modeling processes. In this sense, the need for tools capable of integrating multiple sources of semantic meaning for morphological situations becomes evident. Conceptually, this work aims to continue integrating the process with formal-domain ontologies. It aligns with the B-IRO model, and since the structure of RDF triples follows a pattern like IRO (Input, Relationship, Output), it is possible to integrate the B-IRO model introduced by Jiao et al. (2023) with ontologies formed by triples. The planned flights were recorded in specific ontologies within the maintenance domain, considering the creation of an incremental knowledge base, composed of all drone inspections performed throughout the asset’s life cycle. This process is carried out using the ontological constructor (Menegotto, 2025), for which the set of classes, object properties, and data properties was expanded, enabling the generation of ontological instances that represent the concrete facts and situations observed during the inspections. These instances are produced in OWL and TTL formats. Each flight performed is documented in an Excel file specifically structured for the automatic creation of the ontology, which must later be published on a semantic web platform to ensure accessibility, interoperability, and long-term knowledge integration.
The file Ontologia_Propriedades.xlsx contains the complete structure of object properties and data properties used across the entire ontological domain complex. Its primary function is to organize and centralize all properties created for each structured domain ontology, ensuring alignment with other domains, consistency and reuse. This does not imply that every ontology must incorporate all available properties. Each domain selectively adopts only the properties that are relevant to its scope, expanding its set as new knowledge emerges. Additional properties may be introduced whenever it becomes necessary to declare new factual knowledge about a given topic. Table 1 presents a subset of the properties declared in the file that are specifically intended for building inspection purposes. Column F functions as a thesaurus of actions, grouping the expressions in Column G into affinity categories.
The object properties (Columns C and D) are designed to be congruent with the data properties, a distinctive feature of the ontology builder: for every data property, there is always a corresponding object property. These object properties are generated automatically through formulas that use the contents of Columns F and G. The formula replaces the prefix ‘o.’(meaning of) with ‘f.’ (meaning for) in the action-group identifier and adds the prefix ‘is.’ to the property name, while preserving its semantic core. Column G also includes a validation rule that checks duplicate values, helping prevent the creation of redundant or conflicting properties.
Tables 2 and 3 contain the list of factual events, represented by individuals registered according to the classes and properties declared in a maintenance ontology. The Revit programmed interface (API) for this experience is shown in Figure 4.
Below, two OWL instances written by the Ontology Constructor v.5.0 are presented as RDF resources formatted in Turtle syntax. They represent two Individuals: a registered flight plan and a completed flight, with their spatiotemporal record declared.
-
bim:Plan02 rdf:type owl:NamedIndividual , bim:Flight.Plan ;
-
bim:is.drone bim:DJI001 ;
-
bim:description "It is an autonomous planned flight for Mavic 2 drone" ;
-
bim:description "It is an autonomous planned flight plan for CT of UFRJ" .
-
-
bim:Flight02 rdf:type owl:NamedIndividual , bim:Autonomous.Flight;
-
bim:is.drone bim:DJI001 ;
-
bim:is.completed.flight bim:Fligh02;
-
bim:is.date bim:Date02 ;
-
bim:description " It’s an autonomous planned flight completed over UFRJ CT building";
-
bim:kml.format ”VG_CT_Principal_2024_CT_Roof_A_A_2024_08_08_08_18.kml";
-
bim:is.belongs.to bim:BankIma02 .
-
4 Results and discussion
The ontology results of the methodology presented, together with the GitHub repository containing the Excel files with the structured contents, can be accessed at the following address:
On that page, there are three HTML interfaces designed to allow navigation through the properties used by the ontologies, the entities of the created ontology, and the dynamic graphs for querying the ontology’s entities. To consult the ontology designed for autonomous drone flights case, the VANT folder within each HTML interface can be accessed. This will display either a textual list of classes and instances for consultation or the corresponding graph.
In the graph presented in Figure 5, green arrows indicate the relationships of an individual (magenta dots) with its associated data property values, while blue arrows represent relationships between individuals through object properties. Dashed gray lines show relationships between classes (golden dots) and individuals.
4.1 The OWL ontic hypothesis
Philosophy teaches us that, from its point of view, there is a difference in order between the ontological notion and the ontic notion (Maciel Lima; Reis Pereira, 2024). The first concept is of a higher order than the second, since philosophy considers that Being is that which manifests itself in the multiplicity of the becoming of beings (entities). Based on this premise about the philosophical and hierarchical difference between Being and beings, it becomes possible to think of an architectural project as the construction of conceptual and factual statements in the ontic level about beings or entities, since it is evident that an architectural project will hardly manifest itself to us as an essence. Being, in principle, something nonexistent, the project gradually comes into existence; therefore, its manifestation as Being is that of a constant and sustained becoming in time. In other words, architecture is made of works whose essence results from the confluence of decisions taken in a continuous temporal becoming. Ontological statements would be impossible to define a priori. Their factual evidence will always be expressed at the ontic level: descriptions of concepts and facts that accumulate and extend throughout its life cycle, from its ideation to its demolition, and even more beyond that temporal moment. The abstract level of the Being of an architectural project is definitively close to architects from the outset.
Ontologies applied to architectural design can define a set of class and property declarations (in accordance with Smith’s realism) that participate in the project through the factual concretions of the building, encompassing its urban surroundings, geometric configuration, symbolic load, and other relevant attributes. Each conceptual and factual determination declared in an OWL/TTL file becomes a virtual brick added to the computational representation of the project. Beginning the process with a consolidated design ontology is impossible, as it will always be subject to potential changes. The ontogenetic aspects of a building can only be discerned once the design decisions have been finalized; they cannot be known beforehand. However, this does not prevent guiding the design by relying on prior ontologies that contain descriptive declarations of what is already factually known and accepted, such as typologies of other projects that have endured over time, standards and requirements that must be met, material determinations that define the project, and so on. Nevertheless, all this ontogenetic information still does not define the essence of the project. The individual that is in the process of determination will incorporate its determinations throughout its life cycle, in which the completion of the project is just another link in the long chain of determinations that will continue to expand in its temporal becoming.
Each informational definition incorporated into the BIM project can first be applied to execute construction operations and later leveraged to perform the maintenance tasks highlighted in this article. Drone flights, conceived by the design team, can be planned during the project definition phase, prior to construction. They may be used to carry out inspection operations on the construction site and, in the future, to continue accumulating organized factual knowledge.
But all these subsequent processes need an integrated digital infrastructure. Without sets of factual information about the project elements - such as the construction codes defined by regulatory bodies (DECEA, ANAC, ANATEL, ABNT, IRAM, etc.) - these processes become extremely difficult or even unfeasible.
For this reason, it is argued that the ontic, ontological, or computational representation of a project should not be regarded as a kind of magic formula that, by itself, introduces intelligence into the process, nor as something to be stored and forgotten in dead archives—as once occurred with documentation drawn on drafting film, with the archiving of drawings and models in digital media, or with storage in more recent closed information silos and even closed BIM Common Data Environments.
On the contrary, it is argued that such representation should remain active throughout the building’s life cycle, embedded within an infrastructure of open and continuously maintained networks that incorporate the full functionalities of the Semantic Web and allow knowledge to be progressively integrated over time. If the process is approached from this perspective, and the pretension implied in defining the philosophical universality of a Being is set aside, many of the conceptual doubts that arise for ontologists—and that often hinder the adoption of this technology by immobilizing its creators in the speculative exercise of seeking the singular universality of a Being—could be dispelled.
Hence the boldness in viewing OWL files no longer as definitions of an Ontology Web Language, but rather as a stream of ontic definitions, more accessible as a speculative task and as a form of knowledge articulation. The task, then, would be to understand OWL as a form of Ontic Web Language, according to the hypothesis introduced here. Only over the course of its useful life might a building, designed and described through its conceptual and factual determinations, perhaps, and only perhaps, come to be universally acknowledged as a Being, in the philosophical sense of the term. For this hypothesis, the definition of data structures for CAD/BIM design systems may need to abandon the ideal of seeking the structure of a so-called “single source of truth”.
From a computational point of view, it may be said that a building will always be the superior and complex particular entity of a system composed of a set of lower, particular entities of decreasing and mutable complexities. In buildings conceived in the 1960s, 70s, or 80s that must undergo retrofit interventions to adapt to current technological systems, the problem faced by the contemporary architect is deciding how to respect the ontogenetic reasoning of the original project conceived by its author. A question then emerges: how can the ontogenetic reasons implicit in the work be communicated? Today, this depends on the interpretative ability of the architect, who can grasp these reasons both through direct observation of the work and through its documentation. Maintaining an explicit ontic description of the building can help make these ontogenetic reasons clearer, enhancing interpretability as time progresses, that is, preserving the memory of processes and the explicit knowledge of the reasons behind decisions. In this regard, the conclusion of Bajwa et al. (2024) is particularly relevant, as it emphasizes that model interpretability transcends a mere technical preference; it constitutes a fundamental prerequisite for the effective deployment of AI-based techniques and diagnostic systems in building operations. In order to furnish AI agents with consistent representations of a building’s maintenance history linked to its formal determinations, the establishment of a temporally sustained knowledge base that conserves the memory of processes emerges as a necessity.
4.2 The second challenge: the cognitive infrastructure
Although ontology-building techniques are a specialization within Artificial Intelligence, the work deliberately avoided focusing on the concept of ‘intelligence’, opting instead to emphasize the notions of ‘knowledge’ and ‘knowledge representation’ as preliminary and necessary stages for constructing systems intended to exhibit intelligent behavior. The work advanced toward the description of conceptual and factual knowledge related to building entities. In other words, each formalization of an ontology will always be a descriptive exercise of instances that must be maintained over a period characterized by constant transformations. Although generative AI systems have developed impressive capabilities, the architecture of a more reliable and integrative AI system is still a promise. Yet, with the resources already available, the design of built environments continues, but there is a lack of factual knowledge base for concrete architectural processes. In this sense, it is affirmed that the development of a digital infrastructure network fully integrated into the Semantic Web paradigm, with associated cognitive functionalities, is necessary. Such functionalities, rather than attempting to imitate human intelligence, contribute to enhancing and sharing knowledge about buildings and their surroundings, thereby organizing the factual knowledge of the building life cycle. Intelligence may perhaps find favorable conditions to emerge from a combined form of action between minds, bodies, and working tools, which today undeniably already carry embedded cognitive and computational functionalities. This infrastructure could promote interactions between physical and symbolic contexts, with the capacity to adapt to the multitude of bodily and social beings. In the social community of BIM there is concern regarding open BIM and the development of projects in the IFC standard. In this article, it is argued that such concern is important but insufficient, and that it should be extended to the concept of Open and Linked BIM, through which more semantically ordered data structures could be integrated.
For this reason, thinking about the integrated development of the semantic paradigm of the Internet, one capable of hosting digitally formalized knowledge infrastructure, is a real necessity. Such open networks do not exist yet.
Although it is technically possible for individuals to build them, it is argued that assembling and maintaining a democratic cognitive ecosystem of industrial knowledge — one that allows publicly controlled governance— is a political decision that must involve broader communities willing to assume the responsibility of conceiving and organizing such a cognitive infrastructure, open, shared, and interconnected, according to the paradigm proposed by Tim Berners-Lee (2009). This task, which appears necessary from now on, should be conceived as a cognitive infrastructure capable of growing and sustaining itself over time, given that becoming is an inherent reality of the built architectural and urban environment.
Nowadays, the declarative ontologies with predicates of established knowledge (Xiao et al., 2019) are being integrated into the architecture of generative AI systems modeled with techniques known by the acronym RAG (Retrieval-Augmented Generation) (Zhao, 2024). In this field, developers have promoted a convergence process between AI based on vector NLP systems used in Large Language Models (LLMs) and traditional symbolic Artificial Intelligence. Symbolic AI, formalized through ontology-based techniques, contributes to the process by providing computational representations of factual knowledge about reality. Thus, the use of intelligent agents with greater autonomy is summarized under the term Agentic AI, and the convergence between vector-based and symbolic systems is referred to as neurosymbolic AI (Franz Inc, 2025). It is argued that all CAD/BIM processes need to be integrated with several associated technologies, including:
-
the sociotechnical principles of decentralized data mesh, introduced in 2019 by Dehghani (2023);
-
the application development approach based on Domain-Driven Design principles introduced by Evans (2010);
-
the traditional Linked Data technology for open and connected data proposed by Berners-Lee (2009); and
-
the most recent AI techniques applied to agent-based systems.
In an ideal scenario, a well-designed data-driven knowledge infrastructure (Data Mesh) would need to adopt Linked Data principles to ensure that distributed data are semantic, interoperable, and accessible through open standards. Within such a controlled infrastructure, ontologies can be integrated into mesh-like structures, where each node may be created and maintained. The present discussion highlights the need for agencies such as ANAC, DECEA, and ANATEL to integrate the knowledge they hold into this infrastructure, formed by open and public networks of factual, extensible, and evolving knowledge. Regulations concerning drone flights could be incorporated into this infrastructure of explicitly declared knowledge. Design applications could then rely on certified regulatory content translated into OWL/TTL to apply the appropriate safety parameters for planned flights.
5 Conclusions
The methodology developed in this work has sought to address the difficulties identified by Peinado et al. (2025) regarding the lack of integration among Industry 4.0 technologies. This lack of integration relates to the two challenges identified, for which strategies were proposed to overcome them. The first concerns the creation of ontologies; the second refers to the semantic web infrastructure required to host and enable the combined use of these ontologies, which constitutes the environment where they can interact with each other.
For the first challenge, the contribution lies in the ontology constructor, which highlighted several theoretical concerns related to ontology conception. In the field of architecture and engineering, it is considered a theoretical impossibility to conceive ontologies in terms of the philosophical concept of Being. Instead, ontologists are urged to operate at a more basic conceptual level, centered on the philosophical notion of the ontic—in other words, on objects and entities—consistent with Barry Smith’s realism perspective (Smith, 2004).
The second challenge is more complex, as it involves a socio-political discussion regarding the construction of a semantic web as a cognitive digital infrastructure, capable of enabling the complete integration of processes managed both by human agents and by more recent AI agents. In this sense, certain issues are identified as requiring improvement, with particular attention drawn to the definition of a semantic web network capable of incorporating ontologies as a means of structuring data processes for the maintenance of built assets.
The API developed in this research aims to integrate multiple digital technologies applied to asset management (drones, BIM models, ontologies), beginning in the early stages of the building life cycle. When the BIM model is conceived during the design phase, the next phases can already benefit from flights plan based on the semantics established at this initial stage. In the future, maintenance teams will benefit from the accumulated imagery history while incrementally enriching the building’s records through newly inspected facts, all within a semantic and systematic structure that evolves throughout the life cycle. All these processes require a structured approach capable of supporting long-term, sustainable organization. In this context, it is argued that RDF, when combined with well-defined ontologies, provides a suitable and robust methodological foundation for ensuring consistency, interoperability, and semantic clarity over time.
Accuracy in the geographic positioning of BIM models constitutes a critical factor in ensuring safe drone operations. Inaccurate coordinates or rotational misalignments may result in hazardous flight paths, thereby endangering the integrity of the drone, the built environment, and, most importantly, human safety. Regulatory frameworks established by ANAC (Brazil, 2017) or DECEA (2023) stipulate operational requirements, including minimum separation distances between drones and structures, as well as datasheets with technical specifications for UAS. At present, the application lacks automated mechanisms for retrieving these regulatory parameters. Operators are required to manually input setback distances, flight radii, and altitude thresholds into the API interface, in accordance with the safety requirements applicable to the operational zone. This limitation highlights the second challenge mentioned in this work: the absence of cognitive infrastructure. Regulatory data are often published in unstructured formats, such as PDF documents, which limit interoperability and hinders automation. However, this deficiency also represents an opportunity for technological innovation and the development of new professional roles. In case regulatory authorities adopt a published ontology encompassing all standardized parameters, it would be possible to implement API functions capable of automatically accessing, validating, and certifying updated safe flight data.
It is also important to note that, in autonomous flight plans, an adequate setback distance may still be insufficient if the initial modeling rotation is inaccurate. Therefore, calibrating the model’s position and defining its geographic rotation must be considered critical and fundamental tasks, even if the model has already been positioned and geolocated previously. An angular deviation of just a few degrees can cause significant problems, disrupting the parallelism between the drone’s trajectory and the façade surface, particularly during flights along façades with long horizontal spans, such as those of the Technological Center building. It is essential to emphasize that, for the integration of these various technologies to succeed and enable continuous improvement of management processes, ongoing training of the teams involved is required. Throughout the life cycle of both the asset and the system, teams must continually update their knowledge to ensure proper system functioning. Developing expertise in BIM technologies, Semantic Web concepts, and emerging interfaces is therefore indispensable. The KML files generated by the application can be opened in Google Earth and in the DJI Pilot mobile app. Example KML flight plans can be accessed at:
Future research will continue advancing in semantic modeling through the definition of domain ontologies associated with building maintenance, using RDF triple structures to link external data (such as suppliers) with data related to the operation of the managed asset, including functions, human resources, and service requests from third parties.
Declaration of Generative AI and AI-Assisted Technologies in the Writing Process
Financial Support
Data Availability Statement
All data generated or analyzed during this study are included in this article. No additional datasets were used.
References
- AISH, R. First build your tools. In: INSIDE SmartGeometry: expanding the architectural possibilities of computational design. londres: Wiley, 2013.
- AL-HAKAM, H.; SCHERER, R. J. Integration of BIM-related bridge information in an ontological knowledgebase. In: LINKED DATA IN ARCHITECTURE AND CONSTRUCTION WORKSHOP, 8., Dublin, 2020. Proceedings […] Dublin, 2020.
- ALTINSES, D. et al. Synthetic dataset generation for optimizing multimodal drone delivery systems. Drones, v. 8, n. 12 , p. 724, 2024.
- BAJWA, A et al. A systematic literature review on AI-enabled smart building management systems for energy efficiency and sustainability. American Journal of Scholarly Research and Innovation, v. 3 n. 2, p. 1-27, 2024.
-
BERNERS-LEE, T. Linked Data 2009. Available: https://www.w3.org/DesignIssues/LinkedData.html Access: 15 jul. 2023.
» https://www.w3.org/DesignIssues/LinkedData.html -
BIM4REN. D4.2 BIM4REN repository: software and user guide. Available: https://bim4ren.eu/download/d4-2-bim4ren-repository-software-and-user-guide/ Access: 15 mar. 2024.
» https://bim4ren.eu/download/d4-2-bim4ren-repository-software-and-user-guide/ - BONDUEL, M. et al. The IFC to Linked Building Data Converter: current status. In: LINKED DATA IN ARCHITECTURE AND CONSTRUCTION WORKSHOP, 6., London, 2018. Proceedings […] London: CEUR-WS, 2018.
- BRAZIL. Agência Nacional de Aviação Civil. Regulamento Brasileiro da Aviação Civil Especial – RBAC-E nº 94: Requisitos gerais para aeronaves não tripuladas de uso civil. Aprovado pela Resolução nº 419, de 2 de maio de 2017. Brasília: ANAC, 2017.
- BRYANT, L. R. El principio óntico: boceto de una ontología orientada a objetos. Devenires, v. 18, n. 35, p. 229-264, 2017.
- BUS, N. et al. Towards french smart building code: compliance checking based on semantic rules. In: LINKED DATA IN ARCHITECTURE AND CONSTRUCTION WORKSHOP, 6., London, 2018. Proceedings […] London: CEUR-WS, 2018.
-
DEPARTAMENTO DE CONTROLE DO ESPAÇO AÉREO. Portaria ICA 100-40 Aeronaves não tripuladas e o acesso ao espaço aéreo brasileiro. 2023. Available: https://publicacoes.decea.mil.br/publicacao/ica-100-40 Access: 12 jun. 2025.
» https://publicacoes.decea.mil.br/publicacao/ica-100-40 - DEHGHANI, Z. Data Mesh: entregando valor em escala e orientado a dados. São Paulo: Novatec, 2023.
- EVANS, E. Domain-Driven Design: atacando as complexidades no coração do software. Rio de Janeiro: Alta Books, 2010.
-
FRANZ INC. Agentic AI with AllegroGraph’s neuro-symbolic knowledge graphs White paper. Database Trends and Application, mar. 2025. Available: https://allegrograph.com/wp-content/uploads/2025/03/Powering-GenAI-Apps-with-Knowledge-Graphs-3-2025-2.pdf Access: 12 Jun. 2025.
» https://allegrograph.com/wp-content/uploads/2025/03/Powering-GenAI-Apps-with-Knowledge-Graphs-3-2025-2.pdf - GRUBER, T. R. Toward principles for the design of ontologies used for knowledge sharing. International Journal of Human-Computer Studies, v. 43, p. 907-928, 1995.
- JIAO, Z. et al. A review of theory and application development of intelligent operation methods for large public buildings. Sustainability, v. 15, p. 9680, 2023.
- MACIEL LIMA, J. P.; REIS PEREIRA, C. O princípio de não contradição: princípio ontológico, ôntico e gnosiológico do ser. Revista Opinião Filosófica, v. 15, n. 2, p. 1-13, jul./dez. 2024.
-
MENEGOTTO, J. L. Construtor de ontologias BIM GIS Versão 5.00. Available: https://github.com/JLMenegotto/OntologiaBIM/tree/main/Versão5 Access: 12 jun. 2025.
» https://github.com/JLMenegotto/OntologiaBIM/tree/main/Versão5 - PAN, S. et al. Optimal coverage path planning for UAV-assisted multiple USVs: map modeling and solutions. Drones, v. 9, p. 30, 2025.
- PEINADO, H. S. et al. Detecção de guarda-corpo e rodapés em canteiro de obras utilizando drones e visão computacional. Ambiente Construído,Porto Alegre, v. 25, e137986, 2025.
- POVEDA-VILLALÓN, M.; GARCIA-CASTRO, R. Extending the SAREF ontology for building devices and topology. In: LINKED DATA IN ARCHITECTURE AND CONSTRUCTION WORKSHOP, 6., London, 2018. Proceedings […] Longon, 2018.
- RASMUSSEN, M. H.; HVIID, C. A.; KARLSHOJ, J. Web-based topology queries on a BIM model. In: LINKED DATA IN ARCHITECTURE AND CONSTRUCTION WORKSHOP, 5., Dijon, 2017. Proceedings [...] Dijon, 2017.
- SANTOS, L. R. M. Investigações sobre as primeiras obras de construção da Cidade Universitária da UFRJ. In: SEMINÁRIO IBERO-AMERICANO ARQUITETURA E DOCUMENTAÇÃO, 8., Belo Horizonte, 2024. Anais […] Belo Horizonte, 2024.
- SCHULZ, O.; ORASKARY, J.; BEETZ, J. Lessons learned from designing and using bcfOWL. In: LINKED DATA IN ARCHITECTURE AND CONSTRUCTION, 11., Matera, 2023. Proceedings […] Matera, 2023.
- SMITH, B. Beyond concepts: ontology as reality representation. In: VARZI, A.; VIEU, L. (ed.). INTERNATIONAL CONFERENCE ON FORMAL ONTOLOGY AND INFORMATION SYSTEMS, 114., Turin, 2004. Proceedings […] Turin, 2004.
- SMITH, B. New desiderata for biomedical terminologies. In: MUNN, K.; SMITH, B. (ed.). Applied ontology: an introduction. Frankfurt: Ontos, 2008.
- XIAO, G. et al. Virtual knowledge graphs: an overview of systems and use cases. Data Intelligence, v. 1, p. 201-223, 2019.
- YOUNG, R. I. M. et al. Manufacturing knowledge sharing in PLM: a progression towards the use of heavy weight ontologies. International Journal of Production Research, v. 45, n. 7, p. 1505–1519, abr. 2007.
-
ZHAO, Z. Advanced RAG retrieval strategies using knowledge graphs Generative IA, 2024. Available: https://medium.com/generative-ai/advanced-rag-retrieval-strategies-using-knowledge-graphs-12c9ce54d2da Access: 15 jun. 2024.
» https://medium.com/generative-ai/advanced-rag-retrieval-strategies-using-knowledge-graphs-12c9ce54d2da
