Utilizing Building Automation Systems for Indoor Environmental Quality Optimization: A Review of the Current Literature, Challenges, and Opportunities

Abstract

Indoor Environmental Quality (IEQ) plays a vital role in occupant health and productivity. However, current Building Management Systems (BMS) often struggle in sustaining optimal IEQ levels due to limitations in data management and lack of occupant-centric feedback loops. To address these gaps, this research synthesizes the state-of-the-art methods for IEQ monitoring, assessment, and control within Building Automation Systems (BAS), identifying both technological and methodological advancements, as well as highlighting the challenges and potential opportunities for future innovations. Employing the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) methodology, this multi-stage literature review analyzes 176 publications from 1997 to 2024, with a focus on the decade of rapid technological evolution from 2014 to 2024. The review focuses on high-impact journals indexed in Scopus to ensure quality while acknowledging the potential bias inherent in a single-database search. The synthesis reveals a methodological shift in monitoring from sparse, zone-level sensing towards dense, multi-modal systems that incorporate physiological data via wearables and behavioral recognition through computer vision. Assessment techniques are evolving from static models such as the Predicted Mean Vote (PMV) towards adaptive, personalized frameworks supported by Digital Twins and integrated simulations. Furthermore, control logic is transitioning toward Reinforcement Learning and Model Predictive Control to proactively manage occupancy surges and environmental variables. This evolution of monitoring approaches, assessment techniques, and control strategies is represented within the study’s Three-Tiered Developmental Trajectory, providing a novel Body of Knowledge (BOK) for mapping the transition of building systems from reactive tools to autonomous, occupant-centric agents. This study also introduces a Cross-Modal Interaction Matrix to systematically analyze the systemic trade-offs between IEQ domains. Furthermore, by establishing the “Implementation Frontier,” this work identifies the specific technical and ethical bottlenecks, such as “false vacancy” sensing errors, fragmented data silos, and the ethical complexities of high-resolution data collection that prevent academic innovations from becoming industry standards. To bridge these gaps, we conclude that the next generation of “cognitive buildings” must prioritize three pillars: resolving binary sensing limitations, harmonizing data via vendor-neutral APIs, and adopting privacy-preserving architectures to ensure scalable, interoperable, and occupant-centric optimization.

1. Introduction

The pursuit of a sustainable built environment has evolved from a primary focus on energy conservation to approaches that also prioritize the human experience, as discussed in the study by Zeng and Pishdad [1]. Indoor Environmental Quality (IEQ) has, therefore, become a central performance indicator because it is directly linked to occupant health and productivity, especially in contexts where people spend most of their time indoors. For example, statistics from Klepeis et al. [2] reveals that most people in the United States and Canada spend around 87% of their time indoors. Consequently, buildings must be designed not merely as static shelters but as dynamic systems for human occupancy, management, and operation [3].

This paradigm shift necessitates that the enhancement of IEQ extends beyond the preliminary phases of architectural design and planning to construction and the long-term operational phases of a building’s lifecycle. True sustainability of the built environment encompasses environmental, economic, and social dimensions. While environmental performance has traditionally been measured by energy metrics and carbon footprints, economic performance by operational costs, social performance relies on satisfying occupants’ needs regarding a comfortable and healthy lifestyle. As noted in the study by Zeng and Pishdad [1], social dimension is gaining significance gradually within sustainable development frameworks in the U.S., with the maintenance of excellent IEQ serving as a primary pillar of social sustainability.

1.1. Evolution of Sustainable Building Paradigms

To contextualize the growing significance of IEQ research and its industrial feasibility, sustainable building paradigms are often described as progressing through four stages including: green (energy efficiency and carbon footprint reduction [4,5]), healthy (reducing Sick Building Symptom (SBS) and pollutant exposure), comfortable (maintaining high thermal, acoustic, visual, and air quality standards), and productivity-boosting (supporting cognition, alertness, and decision-making [6]). These stages represent a hierarchy of needs within the built environment (conceptually similar to Maslow’s Hierarchy of Needs [7], see Figure 1).

Figure 1. A hierarchy of needs within the built environment.

To navigate these stages effectively, one of the key approaches the industry has increasingly adopted is Post Occupancy Evaluation (POE), a systematic process for assessing how well buildings meet the needs of their occupants over time [8]. Importantly, when integrated with occupant interface, Building Automation Systems (BAS) serve as the technological engine which makes modern dynamic POE possible. By transitioning POE from static surveys to real-time operations, BAS enables a continuous loop of three major processes: IEQ Monitoring, IEQ Assessment, and IEQ Control. “Static” monitoring leverages established benchmarks from International Organization for Standardization (ISO), Building Management Systems (BMS), and Sustainable Building Rating Tools (SBRT) to verify compliance with predetermined environmental standards and design specifications. “Dynamic” monitoring facilitated by BAS continuously input sensor data and occupant feedback to adapt to changing conditions. Consequently, IEQ assessment integrates both objective environmental data and subjective occupant perceptions, allowing BAS control strategies to responsively optimize IEQ parameters. Overall, the proposed BAS-enabled framework transforms POE from a static retrospective assessment into an active tool. By correlating subjective feedback with objective sensor data, this framework could isolate specific environmental variables that are most critical for IEQ optimization.

1.2. The Role of Building Automation Systems

In parallel with the evolution of sustainability concepts, Building Automation Systems (BAS) have gained ubiquity within the building industry. These systems automate, monitor, and optimize multiple building subsystems, including Heating, Ventilation, and Air Conditioning (HVAC); lighting; mechanical, electrical, and plumbing (MEP); security; and other systems through centralized or distributed control architectures [9]. Historically, BAS architectures were limited to manual oversight or simple rule-based control for complex tasks. The Fourth Industrial Revolution has introduced advancements in Artificial Intelligence (AI) and data analytics that now enable near-real-time analysis and automated decision-making [10]. Several review studies have examined BAS analytics and control, for example as just mentioned AI-big data surveys of building automation and management systems [10] and HVAC adaptive-predictive control reviews [11], while complementary reviews discuss IEQ evaluation approaches such as POE and IEQ assessment surveys [8,12]. However, these reviews are typically subsystem- or method-specific and do not provide an integrated cross-domain synthesis of BAS-enabled monitoring, assessment, and control across thermal, IAQ, visual, and acoustic comfort, which is the focus of this study.

Recent studies have demonstrated that AI-assisted BAS could utilize machine learning models, such as regression, prediction, and classification, to streamline IEQ optimization while addressing energy utilization concerns [10,13,14]. However, many deployed systems still face practical barriers, including limited sensing resolution, fragmented data across proprietary platforms, and weak integration of occupant feedback, which constrain sustained IEQ optimization in real operation. This creates a gap between research potential and operational practice that motivates a structured synthesis of BAS-enabled IEQ monitoring, assessment, and control.

1.3. Research Objectives and Framework

Consequently, this study provides a systematic literature review of the state-of-the-art research focusing on utilizing BAS for IEQ optimization. It offers a comprehensive synthesis of monitoring, assessment, and control methodologies over four distinct tangible IEQ factors: thermal comfort, Indoor Air Quality (IAQ), visual and lighting comfort, as well as acoustic comfort [11]. It also identifies critical gaps in the current literature, specifically addressing the challenges in sensor technology, data interoperability, advanced analytics, occupant-centric dynamic controls, and the management of trade-offs between comfort and energy efficiency. Besides, two major research questions are to be answered throughout the literature review process, which are:

• RQ1: How do Building Automation Systems (BAS) and related technologies currently assist with the tangible monitoring, assessment, and control of diverse IEQ factors?
• RQ2: What are the persistent gaps and technological challenges that hinder the development of fully adaptive and scalable BAS for IEQ optimization?

To address these two research questions, this study employs a systematic literature review following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [12], which encompass a clear database selection, well-defined inclusion and exclusion criteria, and iterative screening procedures to ensure comprehensive coverage of relevant publications. By unifying a well-acknowledged rigorous methodological framework with an in-depth evaluation of articles representing the status quo, the outcomes of this paper include (1) key opportunities in deploying advanced BAS solutions and strategies for future research and practical implementation, and (2) challenges which impede BAS implementation, regarding energy-comfort trade-off analysis, data reliability and security, as well as ethical considerations in building management and occupant data.

2. Research Methodology

To systematically assess the state-of-the-art of IEQ optimization utilizing BAS, and to ensure a rigorous and transparent examination of the existing literature, this study was conducted in compliance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) methodology [12]. The review protocol was not registered.

2.1. Database Selection and Search Strategy

This process began with an initial comprehensive database search aimed at capturing peer-reviewed studies relevant to Indoor Environmental Quality (IEQ), Post-Occupancy Evaluation (POE), and Building Automation Systems (BAS). Academic database Scopus was systematically queried using carefully designed keyword combinations and Boolean operators to maximize coverage. While relying on a single database presents a methodological limitation, Scopus was selected as the primary search engine due to its broad and comprehensive indexing of high-impact technical engineering and building science literature. The decision to prioritize depth within a highly relevant repository over breadth across multiple databases (such as Web of Science or ScienceDirect) was made to maintain a focused, high-quality corpus of core BAS and IEQ research. However, it is acknowledged that this approach limits the retrieved results strictly to publications indexed within this specific engine. To ensure a comprehensive retrieval of relevant literature, the search strategy targeted the core metadata of the literature. Specifically, the query examined the Article Title, Abstract, and Keywords (denoted by the Scopus field code TITLE-ABS-KEY) to capture related terms (see Figure 2). Here are the exact queries inputted in Scopus:

Figure 2. Keyword selection for PRISMA methodology.

TITLE-ABS-KEY (“indoor environmental quality” OR “indoor air quality” OR “ventilation” OR “thermal comfort” OR “lighting design” OR “daylighting” OR “acoustics” OR “noise pollution” OR “airborne contaminants” OR “carbon dioxide” OR “particulate matter” OR “HVAC system *” OR “sustainability” OR “green building *”)

AND

TITLE-ABS-KEY (“post-occupancy evaluation” OR “building performance” OR “occupant satisfaction” OR “user experience” OR “feedback system *” OR “survey *” OR “questionnaire *” OR “LEED certification” OR “zero energy building *” OR “focus group *” OR “user-centered design”)

AND

TITLE-ABS-KEY (“building automation system *” OR “facility management system *” OR “smart building *” OR “energy optimization” OR “energy demand response” OR “sensor *” OR “controller *” OR “actuator *” OR “data logger*” OR “adaptive control*” OR “communication protocol *” OR “smart technology” OR “energy efficiency”)

2.2. Identification of Relevant Articles

This broad search yielded a total of 11,622 documents (see Figure 3 and Figure 4). To focus on contemporary scholarship, the publication window was then limited to the decade preceding the data collection phase (2014–2024), reducing the dataset to 7149 records. This decade aligns with the rapid acceleration of the fourth industrial revolution and the widespread integration of Internet of Things (IoT) sensing technologies into building management systems. Additional filters were then applied to refine the searching results: only peer-reviewed journal articles written in English were included, resulting in a subset of 4033 documents.

Figure 3. PRISMA flow diagram generated using ShinyApp by Haddaway et al. [13].

Figure 4. PRISMA flow diagram created by authors.

2.3. Screening of Relevant Articles

To ensure the review is focusing on multidisciplinary IEQ research and core building science publications, a strict quality control filter was applied to the sources of the publications. To ensure the review focused rigorously on multidisciplinary IEQ research and established building science publications, a frequency-based quality control filter was ap-plied. Journals contributing fewer than 40 articles, equivalent to less than one percent of the search results, were excluded. This threshold was established pragmatically to manage the large volume of initial results and mitigate the inclusion of peripheral or non-standardized sources. While this step successfully refined the dataset to 1133 highly relevant documents, it is acknowledged that this criterion introduces a potential bias by omitting valid studies published in smaller, niche, or newly established journals. In the subsequent stage, the titles and abstracts of these 1133 results were carefully screened for relevance to occupant-centric BAS, IEQ monitoring, assessment, and control, as well as POE methodologies and case studies, leading to 267 results. Only those studies aligned with the scope of this research were advanced to full-text review for final inclusion in the systematic analysis.

2.4. Eligibility of Relevant Articles

During this stage, articles were evaluated and excluded if they (1) lacked a clear Building Automation System control component, (2) focused solely on outdoor environments without indoor implications, or (3) failed to demonstrate an integration of occupant feedback. The selection process utilized a master Excel database where reviewers highlighted the relevance of papers using a color-coding system to facilitate the decision to drop non-relevant studies. As a result, the final screening process yielded 122 articles.

2.5. Inclusion of Relevant Articles

Subsequently, a Backward Snowballing technique, which includes systematically reviewing the references of included articles, identified an additional 54 relevant papers. Thus, the final corpus for the in-depth review comprised 176 articles (see Figure 3 and Figure 4). A comprehensive summary of all the included studies, including bibliographic metadata and research context, is provided in Table A1 in the Appendix A.

2.6. Data Extraction from the Selected Articles

Following the final selection of 176 articles, a data extraction process was conducted to build a comprehensive master database. After categorizing the papers, each author was responsible for extracting the data manually from all papers within their specific category. For each included study, bibliographic metadata, including Authors, Title, Publication Year, Research Location, and Digital Object Identifier (DOI) were extracted and cataloged, facilitating the bibliometric analysis presented in Section 3.1. Beyond general metadata, specific categorical data regarding applied standards, monitoring approaches, assessment methods, and control strategies were manually extracted and tabulated (see Appendix A, Table A2, Table A3, Table A4, Table A5, Table A6, Table A7, Table A8, Table A9, Table A10, Table A11, Table A12, Table A13, Table A14, Table A15, Table A16 and Table A17). This conceptual framework standardized the categorization of monitoring, assessment, and control strategies across all domains, ensuring consistent mapping onto the quantitative distributions. Furthermore, to ensure reproducibility, a subset of the extracted data was independently cross-checked among the authors to verify classification consistency. To visualize the current technological landscape and thematic evolution of these systems, a quantitative content analysis was performed on this extracted dataset. This involved mapping the distribution of research across geographic locations, which is critical in IEQ research as it shows the climatic context and the applicable regulatory standards (ASHRAE, CEN). This mapping identifies whether the current literature is skewed toward specific climates, which affects the generalizability of the findings. Besides, number of publications and building typologies for specific IEQ domains (Thermal, IAQ, Visual, and Acoustic) are displayed in bar charts to illustrate how environmental priorities shift depending on the primary function of the facility. In addition, heat maps are generated to illustrate the distribution of research methods across the IEQ domains. The encoding method for these heat maps utilizes a frequency-based color scale, where each cell represents the absolute count of studies meeting a specific intersection of IEQ domain and methodological approach. Higher color densities (hot spots) indicate established research trends, while lower densities (cool spots) highlight emerging gaps. All bibliometric visualizations and comparative figures were generated using Python 3.12.3 -based data science libraries, utilizing Pandas 2.3.3 for data structuring and Matplotlib 3.9.4 and Seaborn 0.13.2 for the generation of heat maps and bar charts.

3. Literature Review Findings

3.1. Bibliometric Overview and Research Trends

To begin with, a quantitative analysis of the selected literature reveals distinct geographical and typological trends across the four IEQ domains. Geographically, research priorities vary significantly by region (see Figure 5 and manually extracted information from Table A2, Table A6, Table A10 and Table A14 in the Appendix A). Thermal Comfort research is predominantly led by China (22.7%), followed by Spain (12.1%), and the U.S. (9.1%). In contrast, the U.S. (15.8%) is the primary hub for Indoor Air Quality (IAQ), with a secondary concentration of studies emerging from Canada (10.5%), and Poland (5.3%). This distribution suggests that the research priorities are heavily influenced by regional urbanization rates and climate-specific energy mandates. For instance, the dominance of thermal comfort studies in China and the EU correlates with aggressive national targets for building decarbonization and the adoption of high performance HVAC standards.

Figure 5. Research distribution by country for: (a) Thermal Comfort (66 papers); (b) Indoor Air Quality and Ventilation Comfort (38 papers); (c) Visual and Lighting Comfort (14 papers); (d) Acoustic Comfort (4 papers).

In terms of Visual and Lighting Comfort, research contributions are globally distributed throughout North America (USA (14.3%), the Middle East (Egypt (14.3%), Israel (7.1%), Saudi Arabia (7.1%), Iran (7.1%)), and Europe or Asia, rather than a single dominant hub. Acoustic Comfort research remains concentrated in Western nations, specifically Germany (50%), Spain (25%), Portugal (25%)) and the United States (25%), pointing toward a localized maturation of acoustic regulations compared to the more globally harmonized thermal and air quality frameworks.

A temporal analysis of the publication dates indicates a synchronized acceleration in research output across all IEQ domains, particularly following the onset of the global pandemic (refer to manually extracted information from Table A2, Table A6, Table A10, and Table A14 in the Appendix A). This surge reflects a fundamental shift in the “healthy building” paradigm, where IEQ optimization transitioned from a luxury feature to a core component of institutional resilience. Thermal Comfort research has exhibited a consistent annual increase since 2020, reflecting a sustained and growing academic interest in thermal regulation and optimization (see Figure 6a). Similarly, IAQ research shows a clear surge annually from 2021 to 2024 (see Figure 7a), driven by a heightened societal awareness of airborne transmission risks and the subsequent demand for proactive ventilation controls. Visual Comfort displays the most dramatic growth trajectory, with a sharp surge in publications observed in 2023 and peaking in 2024 (see Figure 8a), marking an immediate industry pivot toward human-centric lighting technologies as office environments were redesigned for post-pandemic occupancy. In contrast, Acoustic Comfort follows a more dispersed timeline. While foundational work appears in 2005 and 2010, there is a notable resurgence of interest in 2022 and 2023 (see Figure 9a), pointing toward a renewed focus on acoustic standards in recent years.

Figure 6. Publications on Thermal Comfort per year: (a) stacked bar chart; (b) bar chart showing building types.

Figure 7. Publications on Indoor Air Quality and Ventilation Comfort per year: (a) stacked bar chart; (b) bar chart showing building types.

Figure 8. Publications Visual and Lighting Comfort per year: (a) stacked bar chart; (b) bar chart showing building types.

Figure 9. Publications Acoustic Comfort per year: (a) stacked bar chart; (b) bar chart showing building types.

Regarding building typologies, the literature reveals a divergence in research priorities tailored to the specific functional demands of indoor spaces. Thermal Comfort research is predominantly concentrated on “Educational” and “Residential” environments, underscoring a prioritization of spaces where cognitive performance and restorative sleep are significant (see Figure 6b). The prioritization of living, learning, and working spaces is similar to Acoustic Comfort research (n = 4), where “Educational” environments account for half (2 of 4) of the selected studies (see Figure 9b), underscoring the critical necessity of speech intelligibility and noise control in classrooms. IAQ research targets “Office” buildings as the primary area of attention, followed closely by “Residential” settings (see Figure 7b), reflecting the commercial sector’s focus on air quality as a primary driver of workplace productivity and employee retention. This differs slightly from the thermal comfort literature by placing a stronger relative emphasis on commercial workspaces over educational settings. Visual Comfort research places a predominant concentration on “Office” environments, followed by “Educational” buildings (see Figure 8b), indicating a clear prioritization of visual performance in workspaces where lighting quality is critical for productivity.

The analysis of cited standards reveals differences between globally dominant frameworks and domain-specific national regulations (refer to manually extracted information from Table A2, Table A6, Table A10, and Table A14 in the Appendix A), see Figure 10. ASHRAE Standards (specifically 55 and 62.1) and ISO Standards serve as “common language” across Thermal Comfort, IAQ, and Visual Comfort research, establishing a foundational baseline for global comparison. However, the influence of European standards (EN/CEN) is significantly more pronounced in IAQ and Thermal Comfort studies, reflecting the strong regulatory push for energy efficiency in the EU. Visual Comfort introduces a distinct layer of “Green Certification” influence, with frequent citations of LEED, WELL, and BREEAM alongside technical standards like IESNA. This suggests that visual comfort research is often propelled by market-based sustainability ratings and occupant-wellbeing certifications rather than mere technical compliance. In contrast, Acoustic Comfort operates under a highly fragmented, region-specific regulatory landscape. Unlike the other domains, it relies heavily on national legal statutes (e.g., German Aircraft Noise Act, Spanish legal acoustic requirements) and specific technical protocols like ISO 3382-2, highlighting a reliance on local legal compliance over global consensus and revealing a significant opportunity for the development of internationalized acoustic optimization protocols.

Figure 10. Distribution of standards for four IEQ domains.

3.2. State-of-the-Art in IEQ Monitoring, Assessment, and Control Assisted with BAS and Related Technologies

The integration of Building Automation Systems (BAS) and related technologies such as the Internet of Things (IoT) and Building Information Modeling (BIM), is important for achieving robust monitoring, assessment, and dynamic control of the Indoor Environmental Quality (IEQ) within modern built environments [14,15]. BAS or Building Automated Control Systems (BACS) are defined as computer-based, automated systems designed to manage building’s mechanical and electrical systems. Their fundamental objective is to optimize energy efficiency while maintaining occupants’ comfort [16,17]. The following subsections explain how BAS and advanced control mechanisms address tangible IEQ performance across monitoring, assessment, and control strategies [15,18]. To ensure the stability and reproducibility of this manual classification, the extraction process was strictly guided by the structured taxonomy presented in Figure 11. Monitoring Approaches encompasses all mechanisms used to generate input data, whether physical (sensors), subjective (surveys), or virtual (simulation and archived data). Assessment Techniques refers to the analytical layer where that data is processed against specific criteria. Control Strategies refers to the operational strategies used to modify environmental conditions. Critically, this taxonomy serves as the structural foundation for the bibliometric analysis in this study. The specific sub-domains identified in are standardized to serve as the row labels for the heat map visualizations presented in the subsequent results sections. This ensures that the conceptual classification maps directly onto the quantitative distribution of the literature.

Figure 11. Tree of IEQ representing systematic literature review findings for IEQ Monitoring, Assessment, and Control assisted with BAS and related technologies.

3.2.1. Monitoring Approaches for Four IEQ Domains

Regarding thermal comfort monitoring, both objective and subjective data are fed into the BAS or associated platforms. As for objective data collection, key environmental variables required to calculate conventional thermal comfort indices such as the Predicted Mean Vote (PMV) and Predicted Percentage Dissatisfied (PPD) are measured by sensors [19,20,21,22,23]. Measured key environmental variables include air temperature, relative humidity, air velocity, and mean radiant temperature [19,20,23,24,25,26]. In addition to objective inputs, subjective individual physiological data are collected through wearable sensors [26,27], which include skin temperature (usually measured around hands or neck), heart rate, and metabolic rate [26]. Skin temperature has been demonstrated to be applicable as a predictor for metabolic rate, thermal sensation, and overall thermal comfort [26]. With wrist skin temperature and heart rate data utilized in both linear and non-linear model, predicted thermal sensation could be calculated for integration into control system [26,28]. As for more advanced monitoring techniques than wearable sensors, such as video surveillance, thermal discomfort is inferred by occupants’ thermal-adaptive behaviors [29]. Although this approach avoids the need for wearable sensors and the deployment of an extensive network of in-room environmental sensors, specific human poses or gestures reflecting discomfort status, such as “crossing one’s arms,” “rubbing one’s hands,” or “fanning oneself with an object,” have to be defined clearly [29]. Assisted with computer vision models, these thermal-adaptive actions could be identified and classified within an acceptable precision range [29]. Another benefit for utilizing BAS for thermal comfort monitoring is the ability to visualizing the real-time IoT sensor data within BIM platforms [14]. Facility managers now could interact with a virtual 3D model with monitored thermal conditions, where specific areas experiencing thermal comfort or discomfort levels based on predetermined thresholds could be recorded with time and location. These records could also subsequently generate real-time alarms for facility managers, helping them to make proactive decisions more efficiently [14].

The monitoring of air-pollutant in real-time is increasing vital, especially since occupants spend a majority of their time indoor, where pollutant concentrations are always higher than outdoor levels [30,31,32]. BASs and smart thermostats could offer platforms for extracting and analyzing environmental data, providing opportunities for user-customized ventilation and continuous estimation of air infiltration rates [33]. The emergence of IoT sensors, often categorized as low-cost sensors (LCSs), makes it commercially available to monitor real-time IAQ and deploy control interventions [30,34,35]. As for the kinds of air pollutant that are tracked by those sensors, CO₂, PM_2.5, Total Volatile Organic Compounds (TVOCs), carbon monoxide (CO), Nitrogen Oxides (NOx), formaldehyde, and radon are frequently mentioned within the literature [30,32,34,35,36,37,38,39,40]. CO₂ concentration is widely adopted as an indicator for assessing suitable indoor conditions and triggering ventilation procedures, especially in high-density spaces where human breathing is the primary source of bioeffluents [36,39]. Non-dispersive infrared (NDIR) sensors are usually used for CO₂ monitoring [32], but issues such as sensor drift remain for these commercial CO₂ sensors due to component degradation, aging, and mechanical stress, necessitating frequent maintenance and calibration for precise Demand-Controlled Ventilation [40]. Besides CO₂, VOCs are another major pollutant for monitoring, used by Photoionization Detectors (PIDs) which could be potentially integrated with BASs for real-time ventilation control [30]. With the aid of these PIDs, episodic VOC emission events caused by chemical disinfectants could be identified [30]. Moreover, as for monitoring PM_2.5, consistent accuracy to detect concentration levels remains a challenge for consumer-grade devices [32]. Besides accuracy, the utility of consumer-grade monitors is highly dependent on factors like responsiveness, quality of data visualization, and portability, which directly influence how users interact with the building performance data [32]. In addition to objective sensing, a subset of studies also capture subjective IAQ-related feedback (for example, perceived air quality [41], odor perception, and symptom or complaint reports [34]) to contextualize sensor readings and support decision-making.

The monitoring of visual and lighting conditions relies on collecting comprehensive data regarding lighting quantity, distribution, energy input, and occupant behavior using multiple sensing technologies [42]. A range of sensors, both traditional and advanced, were employed for monitoring visual environments and occupancy in the literature. For example, photometric instruments such as luxmeters (e.g., Sper Scientific Lux/FC PCS-100 840020C and Light Level Sensor PS-2177) and luminance meters (e.g., Konica Minolta LS-110), are utilized for measuring illuminance and point measurements of luminance, respectively. Light intensity sensor modules are strategically placed inside and outside (e.g., 1 m from a window) to detect ambient sunlight intensity [43]. Regarding real-time calibration of luminance values derived from image processing, a chroma meter (e.g., Konica Minolta Chroma meter CS-150) may be used [44]. Besides photometric sensors, modern control systems frequently utilize digital cameras or High Dynamic Range imaging (HDRi) surveillance cameras as HDR vision sensors [44]. Different with simple luminance meters, these HDRi sensors are valued for their low cost and ability to rapidly collect high-resolution luminance distribution data across a wide field of view [44]. To ensure the functionality of the smart light systems, one requirement could be the detection of occupant presence and movement [43]. Common occupancy sensors include passive infrared (PIR) sensors, microwave Doppler sensors, and ultrasound sensor [43]. In conclusion, in terms of values for monitoring, they fall into several key categories necessary for performance evaluation: illuminance (Desk illuminance [45], Average illuminance-Task (E_h-task), and Average illuminance-Room (E_h-room) [42], standardized benchmarks for target illuminance vary based on task complexity, such as 300 lux for general office work and 500 to 1000 lux for more demanding tasks [44,46]), luminance (Desktop luminance (cd/m³), Partition luminance (cd/m³) [45]), glare (Vertical eye illuminance (E_v-eye) [42]), as well as power and energy (Lighting Power Density (LPD) and Energy Use Intensity (EUI), representing the yearly energy consumption per square meter (kWh/m²) encompassing artificial lighting, cooling, and equipment load). Some studies also incorporate subjective visual feedback through POE questionnaires (for example, satisfaction with brightness, glare [47], and lighting quality [48]) as an input alongside sensor data.

The study of acoustic comfort, while critical for optimizing cognitive performance and occupant health, has historically received comparatively less published attention than other core, tangible IEQ factors. This gap is largely due to historical and institutional biases favoring non-educational settings like offices, along with contradictory protocols and inadequate metrics [49]. Current research often overlooks the specific physical and mental health needs of classrooms, prioritizing worker productivity instead. Furthermore, standardized acoustic assessments such as background noise and Reverberation Time frequently mandate closed windows, directly conflicting with ventilation protocols required for IAQ and virus mitigation. Additionally, reliance on single parameters like Reverberation Time in some national regulations is insufficient. Comprehensive assessment must integrate metrics such as the Speech Transmission Index (STI) and background noise levels to capture the full acoustic environment [49]. To address these challenges, within the limited literature, several rigorous methodologies for monitoring acoustic environment has been developed. Acoustic measurements typically employ specialized equipment such as an M2230 microphone connected to an XL2 audio analyzer for signal recording [49]. Additionally, the Nor140 sound level meter (Norsonic) has been specifically used to measure the Speech Transmission Index (STI) via the STIPA method, as well as background noise levels [50]. Parameters under monitoring are quantifying the performance of a space for speech communication. These parameters include Reverberation Time (RT), Background Noise (BN) levels, Speech Transmission Index (STI), and Clarity (C₅₀) [49,50]. Beyond objective measurements, subjective acoustic comfort is also captured in some studies using pre/post occupancy surveys to reflect perceived distraction and satisfaction [51].

Based on the synthesis of the selected literature as well as the manually extracted contents summarized in Table A2, Table A6, Table A10, and Table A14 in the Appendix A, here is a heat map visualizing the monitoring approaches across four IEQ domains (see Figure 12). This heat map reveals a distinct methodological difference between IEQ domains, suggesting that the monitoring protocols are currently domain-specific rather than unified. Thermal comfort emerges as a hybrid discipline, balancing continuous field monitoring with high-frequency subjective surveys, reflecting an understanding that thermal satisfaction is both physical and perceptual. In contrast, IAQ is heavily sensor-centered and relying on CFD simulations, often neglecting the “human-as-a-sensor” feedback loop. This creates an information silo where physiological stress captured in thermal studies is rarely correlated with air pollutant exposure in the same built environment. Furthermore, while Visual Comfort monitoring is increasingly integrated with performance modeling with simulation tools as well as occupancy-sensing via HDRi cameras, acoustic monitoring remains an isolated, episodic activity. The “cool spots” across the het map for wearable sensing and multi-modal data fusion suggest that the industry has yet to achieve a unified occupant profile that captures all four environmental factors simultaneously.

Figure 12. Heatmap of monitoring approaches across IEQ domains. (Note: The data distribution is representative of our specific keyword selection and the papers meeting the inclusion criteria for this review).

To maintain methodological clarity, this study establishes a formal distinction between IEQ monitoring approaches and assessment techniques. Monitoring is defined as the phase of raw data acquisition, encompassing the use of physical sensors (IoT/WSN), subjective feedback (surveys/POE), and virtual or archived datasets (BMS logs/simulations) to generate environmental and behavioral inputs. In contrast, Assessment represents the subsequent analytical layer where these raw data streams are processed into evaluative indices, predictive models, or regulatory benchmarks. In the visual comfort domain specifically, the measurement of point-in-time task illuminance or surface luminance constitutes a monitoring output, whereas the derivation of integrated, climate-based metrics such as Spatial Daylight Autonomy (sDA) or Daylight Glare Probability (DGP) is classified as an assessment output.

3.2.2. Assessment Techniques for Four IEQ Domains

Operational assessment of thermal comfort within BAS focuses on maximizing the thermal satisfaction while simultaneously achieving low energy consumption. Traditionally, PMV and PPD are utilized for determining acceptable thermal conditions [19,52]. However, the reliability of the PMV and PPD model is challenged. Some research shows that it could correctly estimate thermal sensation for only 34% of the cases [53]. That’s why for long-term and holistic assessment of thermal comfort, specifically in dynamic environments, hours of exceedance (frequency of thermal discomfort) [24], exceedance degree hours (summation of the severity of each hour that conditions fall outside a defined comfort zone) [53], and Environmental Quality Index for Thermal Comfort (EQITC) (percentage of time PMV values fall into specific comfort categories) [24], are used to quantify both discomfort severity and frequency.

Considering the assessment approaches for IAQ and ventilation performance, they are increasingly moving forward towards integrated and multicriteria frameworks which combine physical measurements, predictive modeling, and regulatory compliance [54]. Research by Heibati et al. have successfully developed combined simulation tools, including coupling EnergyPlus for energy, CONTAM for contaminant transport, and WUFI for heat and moisture transport, using mechanisms based on the exchange of temperature, airflow, and heating–cooling control variables [54]. Besides integrated models to assess energy, IAQ, and moisture dynamics, simulation techniques, such as CFD analysis and WTT are also integrated to provide essential airflow data for building energy contaminant transport models [55]. In addition to integrated modeling and simulation, several quantitative indices are commonly utilized to measure system effectiveness, including Air Distribution Performance Index (ADPI), Effective Draft Temperature (EDT), Pollutant Removal Effectiveness (PRE), and Infection Probability (IP) [56,57]. Beyond objective assessment, subjective approaches are proved to be crucial, as research by Parhizkar et al. [58] and Anastasi et al. [59] reveal that lower CO₂ concentrations correlate positively with higher well-being indices and reduced reports of respiratory symptom. Subjective feedbacks could also suggest possible perceivable IAQ change thresholds, noting complaints occur more frequently when PM_2.5 concentrations exceed 80 μg/m³ during cooking [34].

Similar to thermal comfort, the assessment of visual and lighting conditions employs a mixed-methods approach integrating objective, simulation-based metrics with subjective evaluations of occupant perceptions. Considering simulation-based assessment, advanced daylighting design usually utilizes dynamic, climate-based daylight simulations (e.g., using engines like Radiance, EnergyPlus, Daysim, Honeybee, and Ladybug Tools) to evaluate performance across an entire year [42,48,55,60]. These dynamic metrics summarize the annual sufficiency and distribution of light [42,60].

Useful Daylight Illuminance (UDI). It measures the percentage of occupied time when daylight illuminance falls within a useful range (e.g., 100–3000 lux [61] or 200–1000 lux [55]), ensuring the space is neither too dark nor excessively bright [62].

• Daylight Autonomy (DA or DLA). It calculates the percentage of occupied hours per year when a minimum illuminance threshold (e.g., 300 lux) is met by natural light alone [42,48,55].
• Spatial Daylight Autonomy (sDA). It evaluates the annual sufficiency of daylight levels by examining the percentage of the working plane area that meets a minimum illuminance level for a specified fraction of operational hours [42,48,63].
• Annual Sunlight Exposure (ASE). It indicates the annual potential for visual discomfort by calculating the percentage of the working plane area that exceeds a high illuminance level (e.g., 1000 lux) for a specified number of hours (e.g., 250 h) [42,48].
• Daylight Glare Probability (DGP). It provides an accurate prediction of the subjective perception of daylighting glare [42,44,48,55]. DGP degrees are classified from Imperceptible (DGP < 0.35) to Intolerable (DGP > 0.45) [44,55].
• Unified Glare Rating (UGR). It is specifically designed for assessing discomfort caused by indoor artificial lighting [44].
• Daylight Glare Index (DGI). It is often used in historical research, though luminance-based metrics are now generally favored [42,60,64].

In terms of subjective assessment, particularly Post-Occupancy Evaluations (POE), questionnaires, often employing Likert scales, are utilized to gather feedback on satisfaction with natural light levels, glare, and lighting quality [42,46,47,48]. These surveys could help determine if luminous conditions meet human needs related to mood, health, safety, and aesthetic judgments [45]. Research from Davoodi et al. suggests that occupants’ perception of the lit environment mainly relies on the instantaneous perception rather than average annual metrics (e.g., DA or ASE) [42]. Furthermore, simple metrics such as horizontal illuminance on the task area (E_h-task) sometimes exhibit a higher correlation with perceived visual comfort than complex metrics like DA or DGP, emphasizing the importance of light quantity on the immediate work plane for user satisfaction [42].

The assessment of acoustic comfort is multifaceted, relying on comparisons of measured objective parameters against regulatory thresholds, scenario-based testing, and increasingly, human-centric evaluation via spatial visualization and subjective surveys. Objective acoustic parameters, including RT and C₅₀, are rigorously calculated based on impulse responses and recorded signals, generally adhering to standards such as ISO 3382–2:2008 [49]. Speech intelligibility is assessed by measuring STI, often using the STIPA method, in accordance with specifications outlined in the IEC 60268–16:2020 standard [49]. Noise levels could be assessed against standards using Noise Criteria (NC) curves, such as NC-40 or less recommended by ASA S.12.2-2008 for open-office plans with forced-air distribution systems [49,50]. Together these metrics characterize reverberation (RT), clarity (C₅₀), speech intelligibility (STI/STIPA), and background noise acceptability (NC). In addition Glean et al., (2022) [50] complementing them with occupant surveys to relate measured performance to perceived distraction and satisfaction.

Based on the synthesis of the selected literature as well as the manually extracted contents summarized in Table A3, Table A7, Table A11, and Table A15 in the Appendix A, here is a heat map visualizing the assessment techniques across four IEQ domains (see Figure 13). Overall, the assessment techniques within IEQ research are divided into prescriptive compliance and predictive modeling. A cross-dimensional comparison of assessment techniques reveals a “maturity gradient” across the IEQ domains. Thermal and visual comfort lead the field in innovation, shifting from static prescriptive compliance (e.g., PMV or 300-lux thresholds) towards dynamic, predictive modeling using Machine Learning and Multi-Objective Optimization. In contrast, IAQ and acoustic assessments remain largely defensive and prescriptive, relying heavily on standards compliance by comparing levels against WHO or ASHRAE limits. Besides, it also focuses more on statistical correlation rather than predictive AI models. This lack of analytical harmony prevents the development of a standardized IEQ communication protocol. Currently, assessment models operate in parallel, failing to account for how perceived comfort in one domain (e.g., visual satisfaction in a daylit space) may broaden the acceptable range in another (e.g., thermal tolerance), a synergy that remains an unexploited opportunity for energy optimization.

Figure 13. Heatmap of assessment techniques across IEQ domains (Note: The data distribution is representative of our specific keyword selection and the papers meeting the inclusion criteria for this review).

3.2.3. Control Strategies for Four IEQ Domains

Moving away from fixed setpoints to dynamic adjustments catered to individual thermal satisfaction, BAS enabled control strategies could help achieve user-centric thermal control. This control integrates personal thermal comfort profiles into the Heating, Ventilation, and Air Conditioning (HVAC) control loop [27,65,66]. Some control systems utilize advanced computation, such as Machine Learning (ML) to forecast thermal comfort states and dynamically adjust system operation [65]. Some systems adopts fuzzy control algorithms, using the calculated group thermal sensation derived from individual thermal sensations to optimize and adjust the zone temperature setpoint [28,67], to balance between comfort and energy savings by setting indoor temperatures near the neutral PMV value [67]. Besides centralized thermal control, utilizing decentralized or localized heating and cooling systems could also optimize thermal environment to occupants’ needs [68]. Examples include heated or cooled office chairs, thermoelectric walls, and fans on the desk [26,68]. The control logic for these decentralized systems is often based on local sensation values of specific body parts such as head, arm, pelvis, or overall thermal sensation values [68]. Collectively implementing adaptive controllers with decentralized HVAC systems could allow a wider acceptable thermal range, leading to significant reductions in energy consumption while maintaining occupants’ comfort.

In terms of control strategies for IAQ, dynamic and energy-efficient management of ventilation and HVAC systems could be facilitated by the advancement of BAS and artificial intelligence. Compared with Constant Air Volume (CAV) systems, Demand-Controlled Ventilation (DCV) systems are a primary strategy for optimizing airflow through dynamically adjusting ventilation rates based on real-time sensor data, resulting in potentially up to 40% energy savings [36]. Among all the DCV systems, CO₂-based DCV is widely researched and implemented, particularly in commercial and educational buildings, due to the link between CO₂ concentration and ventilation needs [36,40]. One challenge for implementing DCV systems in multi-zone buildings is sensor infrastructure costs, which is then addressed using an “adjacent control” method [40]. Regarding the assistance of AI for control strategies, it helps improve traditional rule-based control (RBC) and Proportional-Integral-Derivative (PID) methods by incorporating predictive and adaptive capabilities [69,70]. Among all the advanced control methods, three concepts are mentioned frequently: Model Predictive Control (MPC), Reinforcement Learning (RL), as well as Fuzzy Logic and Neural Networks. MPC, although proved to be an optimal strategy successfully balancing energy usage while maintaining indoor conditions, relies heavily on the quality and accuracy of the underlying thermal and energy simulation models [69,70]. Unlike MPC, Reinforcement Learning (RL) algorithms, often integrated with deep neural networks (DQN), are model-free techniques used to optimize the operation of multiple indoor climate devices (including air purifiers, ventilation systems, and kitchen hoods) to maintain IAQ with energy efficiency [69]. According to the research from Kim and Moon, RL-based controls have been demonstrated to reduce indoor CO₂ concentration and energy usage simultaneously [69]. Besides these two methods, Artificial Neural Networks (ANNs) and fuzzy models (such as the Takagi-Sugeno fuzzy model) are also employed for predicting CO₂ concentration, modeling chiller components, and enabling real-time optimal control schemes in HVAC systems.

As for control strategies for Visual and Lighting Comfort, modern lighting control systems are usually automated, integrating information technology with sensors and actuators [43,46]. These systems, incorporating Internet of Things (IoT), Building Information Modeling (BIM), and Wireless Sensor Networks (WSNs), could automatically regulate the dimming levels of Light Emitting Diodes (LEDs) based on real-time inputs such as available daylight, occupancy presence, and predefined schedules [43]. One specific advanced context-aware control system mentioned by Budhiyanto et al. is High Dynamic Range Image (HDRi) Surveillance Lighting Control System (HSLDCS), which utilizes HDR vision IP cameras to measure luminance distribution and automatically adjust window blinds and LED lamps according to predefined control rules tailored to various activities, such as discussion and slideshows [44]. This HSLDCS implementation has demonstrated 43% to 63% energy savings, depending on daylight availability [44]. Besides, Multi-Objective Optimization (MOO) frameworks utilizing algorithms, like NSGA-III, genetic algorithms (MOGA), and artificial neural networks were also employed in several studies to create a proper trade-off between energy efficiency, daylight, and thermal comfort [46,48,61,63]. Various optimal design variables were identified to help balance competing performance objectives, including orientation, window-to-wall ratio, and shading design parameters [61,62]. Specifically, to ensure the operation of MOO frameworks, shading systems are becoming increasingly dynamic and adaptive to manage solar radiation and glare, particularly in highly luminous climates [62,64]. Three typical dynamic shading devices were employed based on the investigation of the literature, which are movable shading devices, photovoltaic integrated shading devices (PVSDs), and biomimetic adaptive shading skins. According to the studies by Pérez-Carramiñana et al., external movable shading devices such as Venetian blinds or louvers are generally more effective than interior systems regarding a reduction in cooling demand and improvement of thermal comfort, since they could diffuse daylight and achieve a more adequate, homogeneous illuminance [46]. In terms of dynamic PVSDs, it could generate electricity and simultaneously reduce cooling loads [71]. Control strategies for dynamic PVSDs include optimizing the individual and combined movement (e.g., dual-axis rotation) of the panels to balance PV generation, building loads, and daylighting throughout the year [71]. At last, as for biomimetic adaptive shading skins, they were inspired by biological systems such as the Gazania flower or tree morphology. They could dynamically adjust their geometry in response to environmental triggers like light and temperature to optimize daylight exposure, reduce visual discomfort, and enhance energy efficiency [63].

The control strategies for maintaining acoustic comfort generally involve passive design and surface treatments, active sound management systems, and measures specifically designed to mitigate the conflict arising from necessary ventilation protocols. Effective control starts with the architectural design and materials selection to acoustically separate different functional areas and significantly increasing sound absorption through the use of high-performance surfaces [50]. Within the specialized subset of open-plan office research, active sound management serves as a targeted control strategy to improve speech privacy. This involves the strategic addition of broadband background noise to raise the ambient sound floor (e.g., NC−40 or 46dBA to 48dBA), thereby reducing the distraction of irrelevant speech [50]. A major challenge for acoustic control, especially in educational buildings that rely on natural ventilation, is balancing acoustic requirements with the necessity of achieving adequate IAQ, particularly during periods requiring increased ventilation to minimize airborne virus transmission. Since opening windows and doors could increase background noise from external sources (such as road traffic or pedestrian noise), which in turn worsens STI and speech intelligibility. Research suggests that strategic choices regarding classroom location and orientation can help maintain background noise levels close to recommended ranges even with windows open [49].

Based on the synthesis of the selected literature as well as the manually extracted contents summarized in Table A4, Table A8, Table A12, and Table A16 in the Appendix A, here is a heat map visualizing the assessment techniques across four IEQ domains (see Figure 14). The control strategies landscape is characterized by a “conflict zone” where optimization in one domain frequently compromises another. Our synthesis identifies a primary operational conflict between IAQ and acoustics: the increased ventilation rates required for virus mitigation and air freshness often raise background noise levels, directly degrading speech transmission index and cognitive performance. While the thermal and visual domains are beginning to bridge this gap through integrated MOO frameworks, such as optimizing dynamic shading to balance solar heat gain and glare, IAQ and acoustics remain operationally isolated. The transition toward “cognitive buildings” is currently hindered by these trade-offs. While thermal and IAQ domains utilize Reinforcement Learning (RL) for autonomous control, they often do so without acoustic-awareness, potentially leading to noise-included dissatisfaction. Future industry requires moving beyond component-level control toward systemic architectures which resolves these cross-dimensional trade-offs in real time.

Figure 14. Heatmap of control strategies across IEQ domains (Note: The data distribution is representative of our specific keyword selection and the papers meeting the inclusion criteria for this review).

3.3. Synthesis: The Developmental Trajectory of IEQ Optimization

The systematic synthesis of the reviewed literature reveals a clear developmental trajectory for BAS in managing IEQ. Rather than progressing at a uniform rate, the four IEQ domains exhibit varying levels of technological and methodological maturity. To provide a comparative framework for this evolution, this study introduces a three-tiered developmental trajectory, demonstrating the transition of building systems from simple, reactive tools toward autonomous, occupant-centric agents.

Tier 1, foundational or the compliance phase, represents the traditional baseline of building management. At this stage, the primary objective is meeting minimum regulatory standards through fixed, “one-size-fits-all” setpoints. Monitoring is characterized by sparse, zone-level sensing often prone to “false vacancy” errors, while assessment relies on static, universal models like the PMV. Control logic remains rooted in manual behaviors or rule-based schedules. Currently, in contrast to the autonomous optimization seen in thermal systems, Acoustic Comfort remains primarily in the “foundational” tier. While specific interventions, like sound-masking, are utilized in niche contexts, the broader domain still relies heavily on passive architectural design and surface sound absorption, with significantly less adoption of real-time, BAS-integrated optimization algorithms compared to the thermal or visual factors.

Tier 2, integrated or the data-driven phase, marks a transition towards continuous field monitoring and digital integration. This tier utilizes IoT-based sensor networks and low-cost sensors to better align building operations with real-time environmental conditions. Assessment techniques evolve towards statistical correlations and integrated multicriteria frameworks, such as coupling contaminant transport with energy models. Control strategies frequently implement Demand-Controlled Ventilation and PID loops that adjust setpoints based on live sensor feedback. Indoor Air Quality is the dominant domain in this tier, having successfully integrated real-time CO₂ and pollutant tracking into dynamic ventilation logic.

Tier 3, Intelligent or the cognitive phase, represents the frontier of the field, where buildings possess “reasoning” capabilities to proactively balance conflicting goals of energy efficiency and individual well-being. Monitoring at this level is dense and multi-modal, incorporating physiological data from wearables and behavioral recognition through computer vision to “see” and “feel” the environment like a human. Assessment utilizes adaptive, personalized frameworks supported by Digital Twins and machine learning to predict individual comfort states. Control logic is fully autonomous, leveraging MPC and RL to manage environmental variables. Thermal and Visual Comfort are currently leading domains migrating into this intelligent tier, utilizing MOO to resolve complex trade-offs in real time.

Here are three tables summarizing the characteristics of monitoring, assessment, and control for each tier, as well as examples of each IEQ factor at each tier, see Table 1, Table 2, and Table 3.

Table 1. Developmental trajectory of IEQ monitoring approaches.

Tier Features	Tiers	Tier 1 Foundational	Tier 2 Integrated	Tier 3 Intelligent
	Tier Focus	Meeting minimum regulatory standards	Real-time monitoring and operational efficiency	Human-centricity and balancing conflicting goals
IEQ Factor	Thermal Comfort	Sparse, zone-level sensors measuring air temperature and humidity	IoT-based sensor networks providing continuous field data	Multi-modal sensing incorporating physiological data via wearables and behavioral recognition via computer vision
	Indoor Air Quality	Periodic manual sampling or fixed carbon dioxide sensors	Low-cost sensor networks for real-time tracking of CO2, PM2.5, and TVOCs	High-resolution sensing capable of isolating personal behavioral information from generic environmental data
	Visual Comfort	Manual point measurements using luxmeters	Integrated photometric sensors and ambient sunlight intensity modules	High Dynamic Range imaging vision sensors and digital cameras for real-time luminance distribution
	Acoustic Comfort	Episodic measurements of background noise and reverberation time	Continuous signal recording using specialized microphones and audio analyzers	Crowdsourced acoustic data and “human-as-a-sensor” feedback loops

Table 2. Developmental trajectory of IEQ assessment techniques.

Tier Features	Tiers	Tier 1 Foundational	Tier 2 Integrated	Tier 3 Intelligent
	Tier Focus	Uses static, universal models and prescriptive thresholds	Employs statistical correlations and simulation-based dynamic metrics	Utilizes adaptive, personalized frameworks supported by Digital Twins and machine learning to predict occupant comfort states
IEQ Factor	Thermal Comfort	Predicted Mean Vote and Predicted Percentage Dissatisfied	Exceedance degree hours and Environmental Quality Index	Adaptive, personalized comfort profiles and Digital Twin-supported predictive modeling
	Indoor Air Quality	Binary compliance with WHO or ASHRAE limit thresholds	Integrated multicriteria frameworks (combining CONTAM and EnergyPlus)	Predictive AI models (LSTM, RNN) for pollutant forecasting and risk-based ethical assessment
	Visual Comfort	Static illuminance thresholds	Climate-based daylight simulations	Real-time luminance mapping and occupant-in-the-loop satisfaction indices
	Acoustic Comfort	Reverberation Time and national legal statutes	Speech Transmission Index and background noise level analysis	Multi-view analytics and automated fault diagnosis in acoustic environments

Table 3. Developmental trajectory of IEQ control strategies.

Tier Features	Tiers	Tier 1 Foundational	Tier 2 Integrated	Tier 3 Intelligent
	Tier Focus	Operational logic is based on manual behaviors or rule-based control with fixed schedules	Implements Demand-Controlled Ventilation and PID control loops that adjust setpoints based on real-time sensor feedback	Leverages advanced algorithms like Model Predictive Control, Reinforcement Learning, and Multi-Objective Optimization to proactively manage occupancy surges and environmental variables
IEQ Factor	Thermal Comfort	Manual adjustments and fixed, schedule-based thermostat settings	Demand-Controlled Ventilation and PID-based setpoint adjustments	Model Predictive Control and Reinforcement Learning for proactive optimization
	Indoor Air Quality	Constant Air Volume and manual window operation	CO2-based Demand-Controlled Ventilation systems	Autonomous AI-driven control of purifiers and ventilation to resolve “false vacancy” gaps
	Visual Comfort	Manual switches and simple, non-dynamic shading	Automated LED dimming based on occupancy and daylight availability	Dynamic, biomimetic adaptive shading and Multi-Objective Optimization for comfort-energy trade-offs
	Acoustic Comfort	Passive architectural design and surface sound absorption treatments	Active sound management to improve speech privacy	Real-time, multi-objective resolution of the ventilation-noise conflict

4. Gaps, Opportunities, and Challenges in BAS-Enabled IEQ Optimization

Despite the advancements detailed above, the systematic review reveals persistent gaps which hinder the development of fully adaptive, occupant-centric Building Automation Systems (BAS). These deficiencies span sensor technology, data interoperability, system integration, advanced analytics, occupant-centric dynamic controls, and the complex trade-offs between comfort, operational efficiency, and energy management. By synthesizing these findings into a unified “Gap → Opportunity → Challenge” framework, this section moves beyond a descriptive summary to propose a targeted research roadmap.

4.1. Theme I. Standardization—From Data Interoperability Silos to Unified Frameworks

4.1.1. The Gap: Connectivity and Integration Barriers

Regarding Sensor and Connectivity Deficiencies, calibration errors could be one of the most observed gaps in the literature. Sensor accuracy is compromised by environmental factors such as lighting and spatial geometry, as demonstrated by Seghezzi et al. [72] in IoT-based occupancy sensors, Unzu et al. [73] in fiber optic networks, and Ding et al. [30] in photoionization detectors for VOC monitoring. Similarly, Pietraru et al. [35] noted calibration complexities in low-cost air quality systems, which are further compounded by limited detection capabilities in specific pollutants or occupancy patterns, with chair-integrated sensors facing reliability issues [74], wearable sensors showing variability in thermal comfort predictions [26,27,75,76], and low-cost IoT submeters exhibiting limitations in real-time fault detection [77,78]. What’s more, high costs and calibration complexities limit sensor deployment in resource-constrained settings, as highlighted by Zhang et al. [79] in their sensor cost-effectiveness analysis, Pietraru et al. [35] in low-cost air quality monitoring, and Fissore et al. [78] in multi-sensor IEQ devices. Beyond hardware limitations, a more systematic issue exists at the network level. Protocol heterogeneity could compromise effective data transmission throughout different platforms, as diverse protocols like BACnet, KNX, and oBIX hinder seamless data exchange, as noted by Neugschwandtner et al. [80], Fernbach et al. [81], and Domingues et al. [82] who concluded that no single approach fully covers BAS functionalities. Mahalik and Lee [83] identified similar challenges with LonWorksTM networks. Moreover, the industry’s shift towards cloud-based platforms introduce latency that impairs real-time IEQ control, as observed by Bharany et al. [84] in energy-efficient systems, Uzair et al. [85] in MQTT-based solutions, and Devagiri et al. [86] in multi-view clustering. Last but not least, IoT data collection raises privacy concerns, with Bakar et al. [87] highlighting risks in occupant data handling, Dibley et al. [88] noting challenges in ontology-based semantic interoperability, and Sovacool et al. [89] discussing demographic differences in adoption. In terms of System Integration Barriers, large-scale BAS deployment faces scalability challenges, with edge computing showing promise but lacking standardized frameworks, as per Habib et al. [18]. Tang et al. [90] emphasized integration issues with BIM and BACnet, while Ramsauer et al. [91] highlighted WSN and occupant feedback integration. Also, the compatibility issues between systems requires more focuses, as seen in EnOcean-based demand response [92], cyber-physical architectures [93], and Domingues et al. noting fragmented BAS functionalities [82]. Lastly, fault tolerance and maintenance is also one of the challenges. Actuator faults in VAV units require adaptive estimation [94], while incomplete sensor data necessitate advanced reconstruction methods, as shown by Zhao et al. [95] and Liang et al. [96] in KPCA-based diagnostics. These technological fragments all contribute to “information silos” that prevent large-scale, scalable BAS deployment.

4.1.2. The Opportunity: Enhanced Interoperability and Data Standardization

The systematic review of these multifaceted connectivity deficiencies underscores a critical need for a paradigm shift from individual hardware fixes towards service-oriented architectures. The literature suggests that the resolution to protocol fragmentation lies in unified semantic models. Establishing standardized communication protocols is critical to ensure seamless data exchange across the building lifecycle [80,81,90].

Recent case studies from the literature illustrate the practical implementation of this opportunity, providing detailed scalability analysis and a critical evaluation of the technical issues that remain. For example, Fernbach et al. [81] proposed a technical framework to map the BACnet interworking model to the OPC Unified Architecture (OPC UA), aiming to provide standardized, technology-independent access to BAS applications. By creating a unified information model, the study demonstrated how OPC UA clients could interact with heterogeneous BACnet devices without the need for technology-specific drivers [81]. Similarly, Tang et al. [90] developed a Model View Definition (MVD) for BACnet within the Industry Foundation Classes (IFC) framework. It utilized Information Delivery Manual (IDM) and MVD methodologies to define an IFC subset schema, which was tested with Revit and a web browser to demonstrate practical BAS information exchange. This study validated the feasibility of integrating BAS data into BIM tools to support lifecycle management [90]. These advancements represent a pathway towards a vendor-neutral digital ecosystem.

4.1.3. The Implementation Frontier: Technical and Methodological Gaps

While these case studies validate the feasibility of a unified framework, their practical scaling reveals secondary layer of “residual gaps” that form the current research frontier, where the focus shifts from “if” these systems can connect to “how efficiently” they can operate at scale. Moving from architectural concepts to industry standards requires addressing:

• Information Modeling and Management Silos. Current BAS management levels still predominantly rely on technology-specific drivers, which prevents a truly technology-independent method of communication between heterogeneous devices. Overcoming these silos requires generic information modeling concepts and the convergence of standards like BACnet into service-oriented architectures like OPC UA to enable well-defined, cross-platform client access [81].
• Protocol Universality and Data Complexity. The current mapping scope is restricted exclusively to BACnet and OPC UA, neglecting other major industry protocols like KNX or LonWorks, which limits the universality of the mapping concept. Furthermore, mapping specific BACnet properties that use complex data types remains a challenge, as they are not yet fully representable in generic OPC UA information models [81].
• Performance Benchmarking and Scalability. Existing research focuses primarily on architectural feasibility rather than providing empirical data on communication latency or throughput overhead introduced by the OPC UA server layer. Scaling to large-scale buildings with thousands of data points requires a deeper investigation into the overhead of maintaining complex address spaces [81].
• Management Level Isolation. The proposed solution addresses interoperability at the management level but does not solve the underlying hardware-level communication barriers between heterogeneous field-level sensors and actuators. True scalability necessitates vendor-neutral gateways [81].
• Semantic Mapping and Structural Constraints. Most existing data mapping between BACnet and IFC is achieved manually, and the validation of exported data remains an inefficient, error-prone manual process. Transitioning to a scalable standard requires the development of automatic MVD-based data validation and vendor-neutral gateways to replace these manual workflows and mitigate data quality issues. Also, existing IFC entities fail to satisfy all BACnet object types, often necessitating the use of generic proxies like IfcBuildingElementProxy, which fail to represent complex object connectivity. Future deployment requires high-resolution entity definitions that can resolve these structural limitations without compromising model integrity.
• Real-Time Data and Tool Interoperability. While the IFC framework is suitable for static metadata, it is not currently considered an appropriate framework for representing the dynamic, time-series data generated by real-time sensors. Furthermore, prominent facility management tools, such as Metasys and Niagara, are not BIM-based, creating significant interoperability barriers that hinder the adoption of standardized IFC data formats in actual building operations.

4.2. Theme II. Intelligence—Scaling Advanced Analytics for Proactive Maintenance

4.2.1. The Gap: Computational and Modeling Limitations

The deployment of intelligent algorithms is currently constrained by a disconnect between high-fidelity modeling and real-world resource availability. MPC requires intensive resources and accurate building models, leading to high computational demands, limiting the usage in smaller buildings, as noted by Hu and Karava [97], De Coninck and Helsen [98], and Gholamzadehmir et al. [15], who reported inconsistent energy savings. Besides, ML algorithms, such as ANNs used by Yang et al. [99], Chegari et al. [100], and Yue et al. [101], require large datasets, posing challenges related to Data Volume and Scalability challenges. Smarra et al. [102] also highlighted interpretability issues in data-driven predictive control. Additionally, variability and inconsistency in reported energy savings, as seen in adaptive-predictive controls [15] and genetic algorithms [103,104,105], necessitates standardized metrics, as emphasized by Salsbury et al. [106] and Schwartz et al. [107]. Furthermore, real-time processing deficiencies persist in the current BAS systems. As noted by Fu and Zhang [70], Kim and Moon [69], and Deng et al. [108], reinforcement learning for HVAC optimization faces latency issues. Besides, inaccuracies in fault diagnosis were reported in the literature. KPCA-based fault detection requires improved feature selection [96], while hybrid models need diverse seasonal data, as shown by Bezyan and Zmeureanu [109], Zhong et al. [110], and Subramaniam et al. [94]. In addition, the functionality of multi-view analytics remains limited. Multi-instance clustering detects deviations but lacks robustness in incomplete datasets, as per Devagiri et al. [86], Xiong et al. [33], and Wickramasinghe et al. [111].

4.2.2. The Opportunity: Advanced Analytics for Predictive Maintenance

To bridge these computational gaps, the literature identifies a shift toward decentralized and privacy-preserving analytics. By moving processing closer to the data source or utilizing advanced machine learning frameworks, researchers are attempting to resolve the conflict between high-fidelity prediction and the latency issues inherent in current BAS. By leveraging machine learning and IoT data streams, building systems can transition from reactive to proactive operation, anticipating maintenance needs and occupant preferences before comfort is compromised. A key development in this area is the “human-as-a-sensor” approach. The recent literature explores this area through privacy-preserving frameworks and wearable-based sensing. For example, Bakar et al. [87] proposed privacy-preserving analytics, utilizing differential privacy, homomorphic encryption, and federated learning, as a primary mitigation strategy. These methods enable predictive maintenance, such as HVAC fault detection, and occupant behavior forecasting while ensuring building-level data processing remains compliant with rigorous regulations like the General Data Protection Regulation (GDPR) [87]. Complementing this study, Salamone et al. [112] highlighted how wearables, such as smartwatches and personal sensors, can collect real-time IEQ data to indicate building-level system failures. For example, localized elevations in CO₂ levels reported by wearables could signal ventilation faults that traditional wall-mounted sensors might miss, prompting BAS to adjust settings accordingly. This “human-as-a-sensor” approach enables more responsive BAS adjustments but necessitates secure data handling to align with ethical privacy considerations [87].

4.2.3. The Implementation Frontier: Sensing Ambiguity and Validation Barriers

However, the logic of these advanced algorithmic approaches remains tethered to the quality and transparency of the underlying data streams. As these models move from theoretical simulation to industry standards, a new set of “Reliability and Ethical Challenges” has emerged:

• Sensing Ambiguity and Privacy Protection. A significant challenge persists in the inability of current sensor networks to differentiate between generic environmental data and sensitive data directly related to occupants. Without high-resolution sensing that can isolate personal behavioral information at the source, privacy-preserving pipelines struggle to effectively target the specific data streams that require protection.
• Empirical Validation of Compliance. While cloud-based platforms and IoT frameworks support real-time data processing, there is a lack of thorough, large-scale evaluation to verify whether these theoretical architectures ensure actual compliance with data protection laws like the GDPR. Current research is heavily skewed toward technical implementation, with only 5% of studies specifically evaluating the effectiveness of existing regulations in the context of smart buildings [87].
• The Calibration Gap. Only 60% of the reviewed cases included an accuracy test against calibrated reference instruments [112]. Reliability for air quality monitoring drops as low as 40%, while acoustic and thermal domains sit at approximately 45%. This lack of rigorous benchmarking introduces significant variability and limits the reliability of uncalibrated hardware for critical BAS integration.
• Open-Source Sharing and Scalability. A critical barrier to scalability is the near-total absence of open-source sharing. Only one of the 68 reviewed studies released their wearable device in the public domain [112]. Encouraging open-source development is essential to promote the transparency, proliferation, and customization of devices across diverse building contexts and software platforms.
• Hardware Integration and Power Constraints. Many prototypes continue to struggle with high power draw from sensors and the need for better integration into clothing and apparel for improved daily usability [112]. Furthermore, using smartphones for acoustic monitoring introduces data inconsistency due to varying microphone sensitivity levels across different hardware models.

4.3. Theme III. Human-Centricity—Occupant-Centric Dynamic Controls

4.3.1. The Gap: Perception and Feedback Modeling

The primary challenge in achieving occupant-centricity is the persistent inaccuracy in detecting human comfort. Video-based systems could achieve only 78% accuracy [29], with wearable devices facing correlation challenges [26,75,76]. Moreover, the reliance on static standards fails to account for seasonal variations and dynamic thermal responses. Li et al. [113] noted weak predictive power of PMV/PPD indices, while Favero [114] highlighted dynamic thermal responses. Also, the feedback integration remains limited. POE surveys reveal satisfaction discrepancies [8,115,116,117], with real-time loops struggling with data ambiguity [118,119,120]. What’s more, since seasonal variations affect comfort zones [52,121], it requires dynamic models which follows beyond static standards [25,67,122,123,124]. This is compounded by responsiveness gaps, where occupancy-based HVAC systems struggle with system latency [125,126] and a lack of effective occupant-in-the-loop integration [127,128].

4.3.2. The Opportunity: Increasingly Occupant-Centric and Adaptive Control

In response to these gaps, the transition from static, “one-size-fits-all” standards to dynamic, occupant-centric models is currently being facilitated by Agent-Based Modeling (ABM) and personalized comfort profiling. These methods allow the BAS to treat occupants as active, intelligent “agents” rather than passive thermal loads, enabling more granular control of the indoor environment. These approaches leverage real-time occupant data and advanced algorithms to dynamically adjust building systems, ensuring personalized comfort while optimizing energy efficiency without the waste associated with static schedules. Experimental evidence supports this trend: Peng et al. [129] developed a learning-based demand-driven cooling control strategy for office buildings, utilizing machine learning to predict occupancy patterns and dynamically adjust HVAC temperature setpoints. The study achieved energy savings ranging from 5.7% (office M1) to 39.4% (office P1) compared to conventionally scheduled cooling systems, maintaining thermal comfort by aligning cooling with actual occupancy. Similarly, simulation frameworks utilizing Agent-based Modeling and EnergyPlus, such as those by Jung and Jazizadeh [66], indicate that integrating personalized thermal comfort profiles into the HVAC control loops could achieve energy savings ranging from −3.5% to 21.4% (depending on occupancy patterns, with positive savings in most scenarios). This demonstrates the potential to maintain or enhance thermal comfort while reducing energy use, effectively addressing economic barriers by optimizing consumption and supporting regulatory frameworks for adaptive control [130].

4.3.3. The Implementation Frontier: Scaling and Fairness Challenges

Despite the significant energy-saving potential demonstrated in these pilots, transitioning these solutions to the complex, multi-zone environments of modern campuses reveal critical “Contextual Friction”:

• The “False Vacancy” Problem. Current binary motion sensors cannot distinguish between an empty room and an occupant sitting near-motionless. For large-scale applications, integrating more sophisticated occupancy-counting or indoor-positioning systems with high-resolution and privacy-preserving will be necessary to accurately infer thermal loads.
• Algorithmic Adaptability. While the K-Nearest Neighbor (KNN) algorithm is computationally efficient, it currently struggles with stochastic, “non-predominant” behaviors, such as short-duration visits or irregular late-day arrivals. Future deployment requires algorithms capable of handling high-variance occupancy patterns without compromising thermal comfort.
• The Density Diminishing Return. Energy-saving potential is inversely correlated with occupancy rates. In high-occupancy zones, individual preferences often converge with generalized models, offering negligible benefits [66].
• Thermal Fairness and Conflict Resolution. In multi-occupancy spaces, applying a single collective setpoint can violate “fairness” if the strategy is biased toward specific individuals, leading to persistent dissatisfaction for others [66]. Transitioning to a scalable standard necessitates robust conflict resolution mechanisms that balance individual preferences with group satisfaction and building-level energy objectives.
• User Engagement and Behavioral Attributes. Current research often relies on restricted engagement models that do not fully account for varied human behavior attributes, such as a user’s willingness to trade off personal comfort for building-level energy saving objectives [66].

4.4. Theme IV. Connection—Resolving Comfort, Efficiency, and Energy Trade-Offs

4.4.1. The Gap: Lack of Standardized Metrics and Comfort Discrepancies

The final systematic challenge lies in the lack of unified frameworks to evaluate the trade-offs between comfort and consumption. Several POE studies reveal performance gaps in green buildings [131,132,133,134,135], with different sustainable rating tools showing different IEQ requirements [134,136,137]. Besides, climate variations could affect energy savings [138,139,140,141], with context-dependent retrofit strategies unequal distribution of hidden energy consumption [142,143,144,145]. Lastly, in terms of comfort-energy discrepancies, sometimes building’s energy consumption does not align with the actual comfort experienced by its occupants. Overheating issues in Passivhaus designs [138,146] and seasonal assessments [52,121,122,147] highlight the need for adaptive strategies [148,149].

4.4.2. The Opportunity: Integrated Recommendations for Trade-Off Management

To resolve these connections, the literature suggest that BAS must move beyond isolated subsystem control towards a multi-objective optimization approach that accounts for the inherent conflicts and synergies between IEQ variables. Addressing these complex interdependencies requires a three-step integrated approach: (1) standardized evaluation frameworks should be developed to establish consistent metrics for assessing these trade-offs [131,132,133,134,135,137], thereby ensuring generalizability across diverse building types and facilitating more reliable comparisons of IEQ and energy performance, (2) occupant engagement tools, such as dashboards and educational platforms, should be promoted to empower users and encourage energy-conscious behaviors [60,115,150,151,152], drawing on insights into dissatisfaction factors and perceptual influences [58,153,154,155,156,157,158,159], and (3) climate-specific adaptive BAS configurations should be designed, incorporating renewable energy sources and adaptive strategies to optimize energy use while maintaining comfort [138,139,140,141,142,143,144,145,148,149], with a focus on life cycle cost analyses to balance long-term sustainability [146,160].

Here is a table representing a systematic analysis of cross-modal IEQ subsystem interactions, see Table 4.

Table 4. Systematic analysis of cross-modal IEQ subsystem interactions.

Primary IEQ Domain	Interaction Nature	Subsystems Involved	Analytical Perspectives and Literature Evidence
Thermal Comfort	Conflict	Thermal Comfort vs. Energy Intensity	Tightening thermal “deadbands” to ensure constant occupant comfort increases HVAC cycling frequency and energy consumption.
Thermal Comfort	Synergy	Thermal Mass vs. HVAC Peak Shaving	Utilizing building envelope thermal mass for night-flush cooling reduces peak daytime thermal demands, allowing for smaller HVAC equipment sizing and better alignment with time-of-use energy rates.
IAQ	Conflict	Ventilation (CO2) vs. Acoustic Comfort	Increasing outdoor air change rates to lower CO2 and VOC concentrations often raises ambient decibel levels via fan noise and duct vibration, forcing a trade-off between respiratory health and acoustic satisfaction.
IAQ	Conflict	IAQ (PM2.5) vs. Energy Efficiency	Implementing high-efficiency filtration (e.g., HEPA) to capture fine particulate matter increases the pressure drop across the Air Handling Unit (AHU), requiring higher fan power and significantly increasing operational energy costs.
IAQ	Synergy	IAQ vs. Cognitive Productivity	Maintaining CO2 levels below 800ppm via occupant-centric demand-controlled ventilation (DCV) aligns energy expenditure with proven improvements in human decision-making and performance.
Visual	Conflict	Visual vs. Thermal (Envelope)	Expanding window-to-wall ratios (WWR) to increase daylighting and views increases solar heat gain and thermal asymmetry, complicating HVAC load management and potentially leading to localized overheating.
Visual	Synergy	Visual vs. Thermal (Lighting)	Transitioning to high-efficacy LED lighting and automated daylight harvesting reduces internal heat gains, providing a lower lighting consumption and a reduced secondary cooling load.
Acoustic	Synergy	Acoustic vs. Privacy (Sound Masking)	Controlled increases in background noise (sound masking) can improve speech privacy in open-office plans, potentially allowing for higher-density occupancy without a corresponding drop in cognitive productivity.
Psychological	Conflict	Automation vs. Personal Control	High-performance MPC optimizes energy but may trigger “reactance” in occupants. The lack of perceived personal control often leads to higher dissatisfaction, regardless of the objective IEQ quality.

4.4.3. The Implementation Frontier: Engagement and Feedback Deficiencies

However, the final barrier to achieving resilient performance is the “Human-Interface Gap.” Even the most mathematically optimized algorithm can fail if the human element, user perception, engagement, and psychological response, is not adequately integrated into the control loop:

• Engagement Barriers: Occupants are often not engaged in managing their energy use, leading to dissatisfaction with building conditions and systems, as highlighted by various surveys [115,153,155,156,161]. When occupants are not active participants in the building’s energy strategy, they are less likely to adopt energy-saving behaviors.
• Interface Complexity: Interfaces (thermostats, control panels) used to manage lighting, heating, and cooling are often complex and confusing, hindering the user’s ability to operate systems efficiently [150,152,157,158,159].
• The Perception-Objective Disconnect: Multi-objective optimization algorithms like NSGA-II could reduce energy consumption but overlook user preferences [61,101,104,105,162], as noted by Roumi et al. [160]. Last but not least, people’s feeling towards their environment is heavily influenced by perception and psychology, not just objective measurements [58,154,156,159].

In conclusion, the synthesis of these four themes underscores the multifaceted and interconnected challenges impeding the full potential of BAS. By prioritizing research in standardized protocols, privacy-preserving analytics, and context-aware interfaces, future advancements could bridge these gaps. Ultimately, resolving these challenges will align BAS development with global sustainability goals, paving the way for intelligent buildings that are both human-centered and environmentally resilient.

5. Conclusions, Limitations, and Future Research Directions

5.1. Synthesis of Findings

This systematic review of 176 peer-reviewed studies confirms that BAS have evolved from passive infrastructure management tools into active, intelligent agents of IEQ optimization. The integration of IoT sensors, machine learning algorithms, and cloud computing has fundamentally expanded the capabilities of BAS monitoring, assessment, and control. This research provides comprehensive answers to the core research questions established in the study’s framework:

• RQ1: Current State of BAS-Enabled IEQ. Regarding monitoring, the field has transitioned from sparse, zone-level monitoring into dense, multi-modal sensing. The inclusion of physiological data via wearables and behavioral data through computer vision allows BAS to “see” and “feel” the environment in ways closer to human perception than traditional thermostatic readings. Assessment techniques show a methodological shift from static, universal metrics (e.g., the PMV model) toward adaptive, dynamic, and personalized assessment frameworks supported by Digital Twins and integrated simulations. Control logic is rapidly moving toward Model Predictive Control (MPC) and Reinforcement Learning (RL), which anticipate future states, such as weather changes or occupancy surges, to flatten energy demand while maintaining excellent IEQ.
• RQ2: Persistent Gaps and Challenges. Despite these advancements, significant implementation gaps remain between academic potential and operational reality. While research focuses on advanced AI and RL, the industry continues to struggle with basic sensor calibration, high capital costs, and a lack of fundamental interoperability. The reliance on disparate protocols creates data silos which hinder large-scale AI implementation. Furthermore, the industry faces significant challenges regarding the ethical dimensions of data collection, requiring a delicate balance between high-resolution utility and the potential risks of privacy leaks.

The synthesis of this review reveals three primary requirements for the next generation of building systems, representing the necessary transition from foundational challenges to transformative opportunities:

• Resolution of the “False Vacancy” Gap. There is a critical need to move beyond binary motion detection to high-resolution, personalized sensing. This is essential to detect sedentary occupants and resolve individual comfort needs that traditional sensors often overlook.
• Harmonization of Data Silos. Future systems must transition from static metadata mapping to dynamic, vendor-neutral APIs. This allows real-world BAS to interact seamlessly with BIM and IoT data streams, fostering operational longevity.
• The Ethics-Performance Balance. Research must prioritize privacy-preserving analytics, such as federated learning and encrypted data handling. This ensures that the “human-as-a-sensor” approach can be implemented without compromising occupant trust or regulatory compliance.

5.2. Limitations of This Review

While this systematic review provides a robust synthesis of BAS for IEQ optimization, several limitations must be acknowledged to contextualize its findings and guide future improvements:

• Temporal and Geographic Scope. The literature search was restricted to English-language, peer-reviewed journal articles from 2014 to 2024, accessed through Scopus. This focus may exclude valuable insights from non-English studies, particularly from regions with unique building practices or extreme climates, such as tropical or arid environments (e.g., region-specific solutions from Asia or Latin America). Furthermore, given that this work reaches the industry in 2026, the rapid acceleration of technologies, specifically generative AI in facility management, occupant-in-the-loop Digital Twins, and privacy-preserving analytics, means that findings from the Scopus data window require continuous updates to remain fully current in this fast-moving field.
• Inclusion Criteria and Source Selection. While this review provides a thorough overview of recent trends, it encounters several limitations. First, the search relied on a predefined keyword strategy applied to specific bibliographic databases. Consequently, relevant studies using different terminology or indexed elsewhere may have been missed, introducing a possible selection bias. Additionally, using predefined technical anchors without the initial use of “vague searching” (e.g., wildcards such as visual*) may have influenced the total volume of articles retrieved. Thus, the results and conclusions of this review are solely based on the included papers, which may affect the generalizability of the findings across different technical contexts. What’s more, the inclusion criteria prioritized English-language peer-reviewed journals and applied a strict frequency-based filter, excluding journals contributing fewer than 40 articles to the initial search. While this threshold ensured a focus on established, high-impact building science literature, it constitutes a significant limitation, as it inherently introduces selection bias. This filter may have inadvertently omitted cutting-edge methodologies or highly specific IEQ case studies published in niche, emerging, or regional journals. Furthermore, the reliance on a single database (Scopus) and the exclusion of “gray literature” and conference proceedings limit the breadth of the review. These exclusions potentially overlook rapid technological advancements in IoT and AI-driven BAS that often appear first in conference venues before reaching journal publication.
• Methodological Heterogeneity. The diversity of methodologies and definitions across the 176 studies complicates direct comparative analysis and generalizability. Significant variations in how thermal comfort is assessed (e.g., PMV/PPD vs. adaptive models) and how BAS architectures or protocols are implemented challenged the synthesis process. This heterogeneity directly impacted the length and formatting of the comprehensive summary tables provided in the Appendix A.
• Non-Technical Barriers. The review primarily emphasizes technical gaps, such as sensor calibration, data interoperability, and computational demands, but provides limited discussion on economic, social, and policy barriers. As highlighted by Pedersen et al. [157] and Sovacool et al. [89], the high initial capital required for high-resolution monitoring and the lack of clear regulatory incentives or policy support often prevent the commercial transition of these integrated systems. User acceptance and institutional readiness remain underexplored areas that are critical for widespread adoption. Additionally, the ethical implications of occupant data collection, such as privacy concerns, are acknowledged but not fully addressed with actionable solutions, constraining the review’s utility in privacy-sensitive contexts.

5.3. Suggestions of Future Research Directions

To bridge the identified gaps and capitalize on the opportunities presented throughout this review, future research should prioritize the following domains:

Standardization of “Human-in-the-Loop” Data. There is an urgent need for a standardized “Occupant-Centric Communication Protocol.” As proposed by Shagar et al. [11], just as BACnet standardizes equipment data, the industry requires a universal schema for subjective occupant feedback to ensure it can be interpreted seamlessly by BAS across diverse vendors. Establishing this protocol is a major industry milestone necessary to move beyond proprietary silos.

Privacy-Preserving AI for Secure Data Collection. Research must move beyond simply collecting more data to collecting data securely. Testing Federated Learning architectures is a priority. This allows for local data processing on edge devices where only model updates, rather than raw personal data, are transmitted to the cloud. Such techniques are vital to resolving the persistent ethical and privacy concerns inherent in “human-as-a-sensor” behavioral monitoring.

Cost-Benefit and Policy Research. Future work must include rigorous economic studies and Cost-Benefit Analysis (CBA) of high-resolution IEQ monitoring. Building managers and stakeholders require empirical, data-driven proof that the energy savings and productivity gains derived from “human-in-the-loop” systems justify the high initial capital investment and potential privacy risks. Investigating policy incentives to offset these implementation costs is essential for commercial adoption.

Cognitive Buildings and Building Cognition. Future work should explore the integration of Digital Twins with cognitive capabilities. A cognitive building would not merely automate tasks but would possess the reasoning power to balance conflicting goals, such as “energy cost reduction” vs. “acoustic privacy maintenance” vs. “flushing airborne pathogens.” Developing robust ethical AI frameworks for these decision-making engines is a critical frontier.

Longitudinal Interoperability and Reliability Studies. While several studies propose interoperability frameworks, such as mapping BACnet to OPC UA [81], few validate these in long-term, real-world deployments. Future research should focus on quantifying the maintenance burden, reliability, and technical debt of these middleware solutions over the entire building lifecycle to ensure they are viable for large-scale portfolios.

5.4. Conclusions

In conclusion, the convergence of BAS and IEQ represents a critical route toward a green, healthy, comfortable, and productivity-boosting built environment. By addressing the specific technical and ethical challenges identified in this review, specifically the implementation gaps regarding sensor calibration and data integration, research can transition from isolated, simulation-based studies to the robust, multi-zone portfolios required for broader impact.

By mitigating the challenges of data integration and placing the occupant at the center of the control loop, the next generation of buildings can move beyond merely being efficient to actively revitalizing for their inhabitants. For building managers and policymakers, the primary takeaway is that interoperability is the prerequisite for intelligence. True IEQ optimization cannot be achieved through advanced algorithms alone until the industry standardizes the human-building interface to allow for seamless, secure, and scalable data exchange.

These findings directly inform the objectives of this study, which aims to bridge the divide between theoretical optimization and the creation of practical, scalable, and occupant-responsive building environments.