A Systematic Literature Review of Occupant-centric Modelling in Building Performance

Buildings are complex socio-technical systems, and their performance is greatly influenced by the occupants who use them. In recent years, occupant-centric building performance modelling has gained popularity to capture the impact of human behavior on energy use, thermal comfort and indoor environmental quality. In such contexts, occupants are heterogenous, and their behaviors are driven by comfort needs, routines, and peer influence which results in a significant variability in building performance. While several studies exist that reviews specific behavioral factors or technologies, there is a lack of a comprehensive, cross-disciplinary review that synthesizes the state-of-the-art in occupant-centric modelling approaches for building. This thesis fills the gap by providing a comprehensive review of the state of the art in occupant-centric modelling, with special focus on multi-agent system (MAS) and Agent-Based Modeling (ABM) representing occupant behavior. A systematic literature review was conducted, supported by bibliometric analysis of 101 peer-reviewed articles following PRISMA guidelines. This bibliometric analysis with VOSviewer software mapped out the structure of the field dividing into five thematic clusters: (1) occupant behavior and building performance, focusing on how dynamic occupant actions influence energy use and indoor climate; (2) integration of digital technologies, integrating occupant information into smart HVAC and lighting control systems; (3) personalized building control using behavioral science, bridging engineering with psychology to tailor building operations to individual preferences; (4) multi-agent intelligent control and optimization techniques, applying multi-agent systems and reinforcement learning to jointly improve energy efficiency and occupant comfort; and (5) social influence on multi-agent decision-modeling in built environment, employing interdisciplinary approaches (e.g., cognitive, social simulations) to capture complex occupant decision processes in buildings. Each cluster is analyzed in terms of central themes, models and theories, methodologies and research gaps. A unified cross-cluster synthesis shows that despite using the term ‘agent’ in Reinforcement learning framework of cluster 4, they lack some ABM features like scenario-based exploration, social dynamics and heterogeneity of agent. Conversely Cluster 5 integrated them but lacks quantitative optimization. Using ABM for scenario design and RL for control can bridge the gap. Building on the insights the thesis suggests some future research directions themed: (1) Real-time adaptive systems (2) multi-scale and cross-domain integration (3) Human-centred interdisciplinary frameworks that reflect social, cultural, and psychological diversity (4) AI-driven and hybrid modelling techniques like reinforcement learning, model predictive control (MPC) (5) Practical implementation through standardization, open-source tools.