In evacuation modeling, architectural configurations are typically treated as fixed constraints, limiting the exploration of adaptive design strategies that respond dynamically to occupants’ movement patterns. This research investigates environment-mediated coordination in a prototypical multi-agent reinforcement learning (MARL) setting. Using Proximal Policy Optimization (PPO), a navigating agent learns escape trajectories, while a design agent modifies door placement solely based on observed movement, without explicit communication. The system was evaluated in a multi-room environment to explore how complementary strategies evolve when navigation performance and immediate floor plan changes are coupled. The results demonstrate successful implicit coordination, with substantial reductions in episode length as the door controller agent progressively shifted door placements towards positions that facilitated faster escape paths. Navigation trajectories converged on paths that maximized the reward of the navigating agent, while door placement strategies minimized evacuation times. Therefore, although allowing architectural changes during episode runtime, the results still show emerging adaptive behaviors of both agents and demonstrate the feasibility of coordinating navigation and design actions through dynamic environment modification. This approach paves the way for adaptive architectural design systems that can dynamically respond to user behavior patterns, with potential applications in performance-based building design tools and in emergency planning optimization.

Recognizing construction activities from images enhances decision-making by providing context-aware insights into project progress, resource allocation, and productivity. However, conventional approaches, such as supervised learning and knowledge-based approach, struggle to generalize to the dynamic nature of construction sites due to limited annotated data and rigid knowledge patterns. To address these limitations, we propose a novel method that integrates Large Language Models (LLMs) with structured domain knowledge via ontology-based prompting. In our approach, visual features such as entities, spatial arrangements, and actions are mapped to predefined concepts in a construction-specific ontology, resulting in symbolic scene representations. In-context learning is employed by constructing prompts that include multiple structured examples, each describing a scenario with its associated activities. By analyzing these ontology-grounded examples, the LLM learns patterns that connect symbolic representations to construction activity labels, enabling generalization to new, unseen scenes. We evaluated the method using GPT-based models on a dataset covering 29 construction activity types. The model achieved an activity recognition accuracy of 73.68 %, and 50.00 % when jointly identifying the activity and its associated entities. Ablation studies confirmed the positive effects of including Chain-of-Thought reasoning, diverse visual concepts, and richer context examples. These results demonstrate the potential of ontology-informed prompting to support scalable and adaptive visual understanding in construction domains.

Selecting an appropriate data persistent system for a specific use case necessitates a thorough examination of the application domain and the characteristics of the data expected to be stored. While comparative studies of data persistent systems exist in various domains, there is a notable absence of such studies concerning building and environmental data management. This research aims to bridge this gap by conducting a comparative evaluation based on building and environmental datasets and use cases. The study primarily focuses on two types of database systems, namely relational database systems and graph-based database systems. Two building and two city models are employed in the evaluation. The building data sets are extracted from IFC models, and environmental data are extracted from CityGML and OpenStreetMap. The assessment involves qualitatively analysing the database design process of the systems and quantitatively evaluating the efficiency of retrieving data from those systems. The comparative evaluation identifies at least two crucial aspects to consider when selecting a suitable data-persistent system for managing building and environmental data. The first aspect pertains to the stability of the data to be stored, along with the complexity of interrelationships within the building and environmental dataset. The second aspect involves the manner in which data is retrieved to accomplish different tasks within the particular business case. The findings demonstrate that use cases that typically manage interrelated data and necessitate the traversal of complex relationships between building and environmental features are better managed by graph-based database systems, particularly when dealing with large datasets. Conversely, relational databases exhibit superior performance for use cases requiring minimal or no relationship traversal, regardless of dataset size. The contributions of this study can serve as valuable input when designing information management tools and systems for building and environmental data management.

Deep Learning (DL)-based visual analytic systems have demonstrated substantial potential in enhancing construction management. Yet, these systems should go beyond merely facilitating recognition-based tasks and strive to capture high-level semantics. While ontologies have emerged as foundational tools for semantic-enriched visual analytic systems, many existing application-oriented ontologies in the construction domain lack a comprehensive conceptualization and formalization of visual information, restricting their robustness and applicability. Furthermore, there is a notable gap in integrating low-level visual information and high-level semantics. To address these challenges, we introduce the Construction Semantic Enrichment ontology (ConSE). ConSE offers a formal representation of visual information relevant to construction site imagery. It is designed to support the development of semantic-enriched visual analytic systems in the construction domain. To evaluate the ontology, we implement an automated consistency-checking process internally. In addition, we execute three task-based use cases to validate the proposed ontology. The evaluation results demonstrate the efficacy and practicality of the ConSE, confirming its ability to bridge the identified research gaps and contribute to the advancement of semantic-enriched visual analytic systems in the construction domain.

Energy-efficient building renovation is an inter-disciplinary task and requires investigation about the building condition in the urban, environmental, and societal context. Existing literature implicitly mentions the effect of surrounding data in different stages of building renovation. Nevertheless, no conceptual framework is available for practitioners to realize the potential of such data in specific phases of the renovation. The main goal of this study is to understand: (1) based on what knowledge framework surrounding geospatial and environmental data can support building renovation projects, (2) if developing an ontology can help representing this knowledge framework, and (3) how experts and engineers involved in the renovation process can contribute to development of this knowledge framework. The results present an ontology that maps surrounding geospatial and environmental concepts for different renovation tasks and use cases within building renovation. The ontology is built upon knowledge captured from previous studies that implicitly mention the effect of these datasets in building renovation, as well as expert knowledge, brainstorming, and monitoring construction sites. Additionally, a semi-structured verification and validation workshop has been performed to incorporate insights from experts directly involved in different stages of building renovation process. This paper contributes to the body of knowledge by generating a common framework for the surrounding data required in building renovation. It has an implication in practice for engineers by providing a shared knowledge framework and for software developers by providing a basis for BIM (Building Information Modeling) and GIS (Geographic Information System) data integration for renovation purposes.

While the use of semantic models has been explored in different areas of the Architecture, Engineering and Construction (AEC) industry to map and formalize knowledge and support different tasks, building renovation seems to be a neglected area in this field. Therefore, this paper presents the Renovation-Installation (Reno-Inst) ontology, which maps knowledge from the renovation domain considering different requirements, constraints, and other elements related to the installation of common renovation products such as windows, thermal insulation panels, and heat radiators. The development of the Reno-Inst ontology relies on an approach combining input from experts and knowledge from related engineering documents. The verification and validation process includes a content evaluation workshop with experts, a verification design stage, and the analysis and implementation of a real case study. The paper provides another example of the power of ontologies as a method for mapping and representing knowledge, gathering, and retrieving relevant information. Particularly, renovation projects encounter diverse challenges that lead to cost and schedule overruns, making their performance typically low. Therefore, the proposed Reno-Inst ontology provides a basis from which new applications can be developed, tested, and deployed to support improvements in the building renovation field, especially in the planning and execution of renovation activities.

Social media provide access to behavioural data at an unprecedented scale and granularity. However, using these data to understand phenomena in a broader population is difficult due to their non-representativeness and the bias of statistical inference tools towards dominant languages and groups. While demographic attribute inference could be used to mitigate such bias, current techniques are almost entirely monolingual and fail to work in a global environment. We address these challenges by combining multilingual demographic inference with post-stratification to create a more representative population sample. To learn demographic attributes, we create a new multimodal deep neural architecture for joint classification of age, gender, and organization-status of social media users that operates in 32 languages. This method substantially outperforms current state of the art while also reducing algorithmic bias. To correct for sampling biases, we propose fully interpretable multilevel regression methods that estimate inclusion probabilities from inferred joint population counts and ground-truth population counts.