Skip to content Skip to navigation

Building for the Occupant: Optimizing Building Layouts for Energy Efficiency and Organizational Performance

Project Team

R. Jain, A. Sonta

Research Overview

Observed Problem:

Managers of organizations and commercial offices are always seeking ways to improve worker performance and reduce energy consumption. Studies have demonstrated methods for improving each goal in isolation, but no research has focused on developing building designs that simultaneously address both.

Primary Research Objective:

We propose a data-driven socio-technical system for facility operations that can create designs and layouts that promote workplace collaboration and productivity while simultaneously reducing energy consumption in commercial buildings. We will deploy smart and connected plug load energy sensors within multiple organizations to develop our methods for occupant activity and network inference. We will also leverage novel data-driven spatial analysis and optimization techniques to evaluate and produce suggestions for new facility designs that will tightly couple occupant behavior, facility management, and organizational operations. Specific research objectives:
  1. Inferring occupant activities
  2. Inferring occupant social/organizational networks
  3. Optimizing layouts for energy and organizational goals

Potential Value to CIFE Members and Practice:

  • The overarching outcome of this work is the development of a data-driven socio-technical system for facility operations that can create designs and layouts that promote workplace collaboration and productivity while simultaneously reducing energy consumption in commercial buildings.
  • This framework will serve as a toolset for commercial building management optimization and re-design. Using simple smart plug load sensor devices—or other similar IoT sensors—a facility manager will be able to conduct a “socio-space audit” of their building to understand how their spaces are being used and their organization is functioning in near real-time. Then, a facility manager will be able to utilize our system to explore and implement a set of optimized layouts designed to reduce energy use and encourage collaboration.

Research provides relevant insights for: owners, operators, and (re)-designers.

Research and Theoretical Contributions

From a theoretical perspective, we expect our contributions to center around new methodologies for inferring spatio-social networks in the built environment from sensor data as well as for optimizing commercial building layouts for energy and organizational goals. In the end, this work aims to enable facility managers to simultaneously fine-tune the performance of their buildings and their most important stakeholders—the occupants in their facilities.

Industry and Academic Partners

While no specific CIFE members were involved in the development of this proposal, we believe there are significant opportunities to leverage CIFE member companies for providing data and testing of our proposed data-driven socio-technical optimization framework for facility operations.  In particular, CIFE member companies who develop design software (e.g., Autodesk) would be beneficial partners for understanding how our proposed methods would integrate with existing modeling and facility design software. Additionally, CIFE member companies who are design-construction-operators (e.g., Ryan Companies) and owner-operators (e.g., Gilead, Google, GSA) of commercial building facilities would be strong partners for further data collection and testing of our proposed framework.

Research Updates & Progress Reports

November 2019 Update

Our research progress to this point has largely focused on the goal of inferring occupant social/organizational networks. For this key step in the research program, we have published a technical letter in IEEE Embedded Systems Letters and have recently completed a full draft of an extended journal article (to be submitted soon). In these papers, we find that our automated method for inferring socio-organizational network structure outperforms other data-driven methods in the literature and matches with network structure measured through survey responses to a statistically significant degree.

We have also developed initial methods for optimizing building layouts, focused for now on energy performance. We presented our initial findings at the International Conference on Applied Energy in August and will focus our efforts now on co-optimizing building layouts to encourage collaboration and reduce energy use.
________________________________________

Detailed Research Overview & Progress Updates

Detailed Research Overview & Progress Updates - April 2020

Because our overarching research objective is to develop a framework for co-optimizing building layouts for both energy efficiency and organizational performance, our work has been focused along these two fronts simultaneously.

Organizational performance: To better understand the relationship between building design and productivity, we have developed a method for accurately inferring the structure of socio-organizational relationships that make up organizations in commercial buildings. Once relationships are inferred, we can analyze the structure of the organizational ties in the context of the spatial layout of the building. In a case study, we have found that the spatial network does not correlate significantly with the organizational structure, presenting the opportunity to better align the human network with the spatial design.

Energy efficiency: The layout of occupants also has an effect on the energy-consuming lighting and HVAC systems within buildings. By monitoring each individual occupant’s use of space in a building, we are able to characterize how similar or different an occupant’s patterns of space-use are to the other occupants near him or her. We are actively working on a framework for grouping occupants with similarities in patterns of space use, with the goal of increasing the alignment between human patterns of space use and building system operation. Our initial results suggest that reorganizing people in space has the potential to save energy.

Below, we discuss the results from each of these research areas in more detail. We also discuss the future steps we plan to take to integrate the two research ideas into an overall framework for building design that addresses both our human and energy systems.

Spatial layout and organizational performance

The structure of organizations can in part be described by the set of relationships among occupants. These relationships include organizational aspects, such as work-related communication, advice, and trust, as well as social aspects, such as friendships. The organizational theory literature has pointed to the understanding of these relationships as foundational to effective management of organizations (Peponis et al. 2007). One component of this management is the spatial layout of the people in physical space, where alignment of the socio-organizational network with the spatial layout is expected to encourage collaboration in the workplace (Kabo et al. 2015). Our overarching goal for this component of the research to develop a method for inferring the structure of organizational relationships, and ultimately understand how well that structure aligns with the spatial layout. The opportunity resulting from this analysis would be to redesign the space to improve this socio-spatial alignment.

In a paper currently under review, we have developed the “Interaction Model” which learns socio-organizational relationships from ambient sensing data. At a high level, this model finds times when occupants have the opportunity to interact with other occupants in the building. Repeated opportunities for interaction are assumed to signal a socio-organizational relationship. Figure 1 shows the overarching process for this model. It begins by abstracting the time series plug load data (Figure 1A) into states of activities (Figure 1B), which describe localized use of space. Leveraging the information embedded in this use of space, the model identifies times when office workers have opportunities for social interaction with other office workers in the physical space of the building (Figure 1C)—these opportunities are the building blocks of the inferred socio-organizational network (Figure 1D).

Because our overarching research objective is to develop a framework for co-optimizing building layouts for both energy efficiency and organizational performance, our work has been focused along these two fronts simultaneously.   Organizational performance: To better understand the relationship between building design and productivity, we have developed a method for accurately inferring the structure of socio-organizational relationships that make up organizations in commercial buildings. Once relationships

Figure 1: Demonstration of Interaction Model steps applied to data from a single day. (a) Raw plug load energy values. (b) Activity states (low, medium, and high energy) resulting from the Bayesian clustering of the energy data. (c) Opportunities for social interaction, where a black value of 1 indicates interaction opportunity. (d) Resulting network (after 365 days of analysis), shown both as an adjacency matrix heatmap and as a graph visualization.

In a case study, we distributed a survey collected ground truth data on social and organizational relationships in an office in Berkeley, CA with 18 members. We also deployed plug load energy sensors and inferred the network using the Interaction Model. Using the Pearson product-moment correlation metric (adapted for networks), as well as the Quadratic Assignment Procedure (QAP) for testing significance of this correlation (Krackhardt 1987), we find that the Interaction Model successfully learns a network that is statistically similar to the ground truth network. Details of the modeling strategy and the statistical tests can be found in our forthcoming paper.

Once we learn the network, a key question is understanding how closely this socio-organizational network aligns with the spatial layout of the building. Using the language of Space Syntax (Bafna 2003), we can abstract the floorplan layout to a graph describing spatial relationships (Figure 2). Once we have the spatial network, we can then test the relationship between the spatial and organizational systems using the same Pearson product-moment correlation. In our case study, we find that this correlation is small and insignificant. This finding demonstrates the potential value in reorganizing the design of the building such that the spatial network aligns with the socio-organizational network, thereby encouraging the activities that define the network (i.e., communication, seeking advice).

Figure 2: Space syntax decomposition of the office in our case study. The theory of Space Syntax defines two relationships: topological depth and angular depth, as indicated in the upper left.

Figure 2: Space syntax decomposition of the office in our case study. The theory of Space Syntax defines two relationships: topological depth and angular depth, as indicated in the upper left.

Spatial layout and energy efficiency of lighting systems

Buildings provide occupants with visual and thermal comfort by consuming energy to operate lighting and HVAC systems. These visual and thermal resources are shared in space. That is, occupants within the same zones are sharing the light and thermal environments created by the systems. Our overarching hypothesis driving this research effort is that by grouping occupants with similar patterns of space use into the same zones, the building systems can better respond to the occupants’ use of space and save energy.

In a case study, we have focused our work on understand the relationship between individual space use and operation of lighting systems. The site for our case study is a single floor of an office building in Redwood City. There are 165 workstations on the floor, and each workstation is outfitted with a plug load sensor recording the aggregated power consumption of computing devices at each workstation at 15-minute intervals. We focus our initial results here on the data from February 2020. The floor has 11 lighting zones (also called motion groups), with each zone providing light to an average of 15 desks (Figure 3). Each fixture has an infrared sensor, and when any fixture within a motion group senses movement, all of the lights within that motion group turn on. If no movement is detected for 5 minutes, all lights in the group turn off. Motion groups that do not serve desks are excluded from the study, since we are interested in characterizing how desk-level occupancy affects lighting energy. The lighting system also includes automated daylight harvesting.

Figure 3: Floorplan showing lighting zones, also known as motion groups (MG).

Figure 3: Floorplan showing lighting zones, also known as motion groups (MG).

Using the power consumption data collected at each workstation, we map the plug load energy to three abstracted activity states, as discussed in our previous work (Sonta et al. 2018). Intuitively, these activity states represent individual use of the workstation, where a high energy state means the occupant is interacting with their workstation, a medium energy state means that some of the equipment has entered a power saving mode, and a low energy state means that all equipment at the workstation is off. Then, using the activity states, we can define schedules for each occupant as the time-series of transitions between the different activity states.

Looking at each lighting zone, a key question is how similar or different are the occupant schedules among the members of each zone? If the schedules are very different, we would expect that to cause an increase in the energy consumption for each zone.

We therefore need a definition of the “difference” or “diversity” among the occupants in each zone. Based on the work in (Yang et al. 2016), we can define this “diversity” as the distance between the vectors of time-series schedules for each occupant in the zone.

Consistent with this previous work, we use the Euclidian distance. Using this distance metric, we can compute the distances between all occupants in a zone, forming a distance matrix. Normalizing the sum of the entries in this distance matrix by the total number of entries in the matrix (except the diagonal), we have an average distance among all the occupant schedules within the zone.

In Figure 4 below, we show the relationships between occupant diversity (as defined by Euclidean distance in schedules) and lighting energy consumption, where each data point refers to data from a single zone over one day. We can see from this figure that there is positive correlation between these two variables, as we would expect (R2 =0.49). Fitting a linear regression model, the parameter describing the change in energy as a result of an increase in occupant diversity is significant (t=15.7, p=0.00). In Figure 4, however, we can see that there are roughly two main clusters in the data. This results from the distinction between working days and non-working days (i.e., weekends, holidays). If we separate out just the working days, as shown in Figure 5, the relationship between occupant diversity and energy becomes less distinguished, but it is still significant in a linear regression (t=5.4, p=0.00). This relationship between diversity in occupant schedules and lighting energy consumption underscores the opportunity for redistributing occupants across the floorplan to reduce diversity and save energy.

Figure 4: Relationship between occupant diversity within the lighting zone and energy consumed by that zone, by day. Data from February 2020.

Figure 4: Relationship between occupant diversity within the lighting zone and energy consumed by that zone, by day. Data from February 2020.

Figure 5: Relationship between occupant diversity and lighting energy, for working days only. Data from February 2020.

Figure 5: Relationship between occupant diversity and lighting energy, for working days only. Data from February 2020.

Integrating energy and work outcomes in design

Based on the work described above, we have evidence that the decision of where occupants sit within a building has an effect on both the energy consumed by building systems as well as the alignment between layout and organizational structure. Reorganizing occupants, therefore, has the potential to improve one or both of these objectives. Our work plan is now focused on creating a framework, based on the evidence discussed above, to optimize building layouts for both objectives:

  1. Develop method for clustering occupants by their similarities in schedules of space-use. This clustering method should reduce the diversity of within-zone occupant schedules, thereby creating the potential to save energy. We are looking into hierarchical clustering, and possibly leveraging the information gained by singular value decomposition.
  2. Constrain the layout of occupants by the structure of the socio-organizational network. By ensuring that the layout of occupants matches the learned relationships, we can create new layouts that are not expected to negatively impact the flow of work within the organization.
  3. Evaluate how new layouts are expected to affect lighting energy consumption. To do this, we are building a data-driven simulation engine that includes the zone-level occupant schedules as a key input. By creating new layouts, new zone-level schedules will be produced, and the simulation engine will determine the expected lighting energy consumption.

Overview & Observed Problem - November 2019

Two of the most important indicators of a building’s performance are its energy efficiency and ability to support the human activities for which it was designed. Researchers and practitioners have made vast progress in understanding each of these indicators in isolation, but few data-driven frameworks have been able to address the inextricable socio-technical link between them.

Imagine you are a manager of an organization in the knowledge industry. Chief among your goals is promoting collaboration among your employees—and thereby, hopefully, the innovative capacity of your organization. Your organization also cares deeply about environmental impact, and as a result your building has energy-efficient controls of heating, cooling, and lighting systems. You are considering redesigning the layout of your office space. It would be best, you think, to leverage your advanced building systems to reduce energy consumption by way of your new layout. But you, like most organizations, value your people and their productivity far more than your building’s energy consumption.  In fact, for the University of California system—whose 2016/17 operating budget is public data—total employee salaries, wages, and benefits were roughly 74 times more expensive than utility bills, underscoring the notion that organizations are rational if they prioritize the productivity of their workforce over energy efficiency. As a result, for you it is integral that you create a new layout that stimulates collaboration and innovation all while reducing energy usage.  For an organization with $74M in personnel costs and $1M in energy costs, a mere 2% improvement in productivity and 10% reduction in energy usage would result in $1.5M+ in added economic value.

We aim to develop a data-driven socio-technical system for facility operations that can create designs and layouts that promote workplace collaboration and productivity while simultaneously reducing energy consumption in commercial buildings (Figure 1).

Theoretical optimization framework.

Theoretical & Practical Points of Departure

Part of the reason for our collective inability to co-optimize for both energy efficiency and building functionality is the lack of a reference point directly attributed to both. How can we measure the productivity impact of improving the insulation of a wall, or correctly sizing a window overhang? How does the addition of more meeting spaces—intended to improve collaboration—directly impact energy consumption? The traditional lenses for viewing each of our two goals are sufficiently different that it has been difficult to simultaneously consider both. It is only recently that the research around energy efficiency has found that the building occupant is the largest and most fundamental driver of building energy consumption (Hong and Lin 2012). This emergence of occupant-driven research opens new opportunities for more holistic building analysis. Our new reference point—the building occupant—brings energy efficiency into the realm of building function.

POD: Occupant-driven energy analysis

A key point of departure upon which our proposed work will build is the availability and analysis of data produced within buildings. In particular, IoT devices deployed at the level of the individual occupant create a wealth of information about users of facilities. Given that the occupant has been shown to be one of the largest drivers of the variability in energy consumption in buildings (Norford et al. 1994), researchers recently have developed data-driven methods and models for understanding occupant behavior (Zhao et al. 2014). Occupant behavior can be described broadly, from simple presence/absence data to occupant adaptive behaviors such as interacting with windows and thermostats (D’Oca et al. 2018). Presence/absence data can be used to derive more accurate occupant schedules, but as Feng et al. (2015) discuss, even this seemingly simple information can vary from the overall building occupancy level, to the occupancy status of a zone, to the individual location of occupants in the building. These kinds of occupant schedules can be valuable for creating more realistic energy simulations, aiding in the design of buildings.

As building systems that control heating, cooling, ventilation, and lighting become more advanced and controllable, contextual information about occupant activities can also be instrumental in improving the operation of buildings. Recent research has demonstrated the potential for human-in-the-loop control of many cyber-physical systems, including HVAC and lighting (D’Oca et al. 2018). By providing these advanced systems with richer information about the occupants’ space utilization across the building, we can ensure they only provide HVAC and lighting systems where they are needed and when they are needed in the building (Dobbs and Hencey 2014).

POD: Spatial effects on occupant performance

Within commercial buildings, occupant-driven research has also recently focused on the performance of occupants in organizations. The study of productivity in organizations is many-faceted, but recent research has noted that space and design can have a large impact on organizational success. Even in the age of digital communication, classic studies have found that spatial proximity is a large predictor of the frequency of communication (Waber et al. 2014). Furthermore, researchers have found that by analyzing the paths in an academic office building, the more likely occupants are to have their paths overlap, the more likely they are to collaborate and be successful in those collaborations (Kabo et al. 2015, 2014).

An area of research at the intersection of organizational behavior and building design has defined a set of tools called space syntax (Bafna 2003). These tools enable building or urban layouts to be defined by the physical barriers and statistically analyzed with regard to connections between spaces. Researchers have found that offices with higher levels of integration and connectivity—space syntax measures related to the ease at which occupants can access other spaces than their own—lead to higher levels of communication and productivity (Congdon et al. 2007). Throughout this area of research, it is common to utilize the network of organizational and social relationships in an organization as input to analysis. Most often, researchers utilize questionnaires or extensive surveys in order to gather this network information, which are both time-intensive and expensive. Research has suggested that automatic inference of such networks could aid our ability to analyze existing organizational layouts and ultimately design more effective ones (Kabo 2018; Sailer and McCulloh 2012).

Ultimately, the research around occupant-driven energy efficiency and organizational success suggests that successful attempts to design and retrofit office spaces should consider both points of view. Recent work suggests that occupant-driven optimization of layouts has the potential to affect energy-efficiency (Yang et al. 2016), but we are aware of no work that seeks to co-optimize these layouts for both energy efficiency and occupant performance. Our dual approach will be crucial to driving change in buildings, as people drive the success of organizations far more than utilities. For an organization where people cost 74 times more than utilities, rational decision-making would prioritize the productivity of the workforce over energy efficiency. While occupant-focused energy interventions have shown potential, managers would be unlikely to make changes if they worry about disruptions to productivity. Our proposed work seeks to embed these organizational constraints—along with energy goals—into the process of redesigning workspaces.

Research Methods & Work Plan

Building upon our two key points of departure through the lens of the occupant—data-driven energy efficiency and organizational performance—we propose to develop a data-driven socio-technical framework for optimizing commercial office layouts for energy efficiency and productivity. Three steps summarize our proposed approach: (1) inferring occupant activities, (2) inferring social and organizational relationships among occupants, and (3) leveraging space-utilization information and the occupant network to optimize layouts. If successful, our proposed work will enable tight coupling of building systems with occupant activities, thus improving energy efficiency. Moreover, it will promote key aspects of productivity in organization (e.g., communication and collaboration), ensuring that organizational managers will be more open to spatial changes within the building.

Inferring occupant activities

Drawing upon our previous work (Sonta et al. 2018), we propose to continue developing methods for accurately inferring activities of occupants across building spaces using embedded sensor devices. As Figure 2 suggests, the raw data from IoT devices—in this case plug load sensor data—do not in themselves provide useful information about the building. We require methods that draw upon the specific engineering context to glean useful information from these data. Our work will build upon a set of statistical clustering tools based on variational Bayesian inference to translate raw plug load sensor data into information about the states of occupant activities. This work is inspired by research in the bioinformatics field that used similar tools to analyze human biological states based on times series heart-rate data (Costa et al. 2012). Similar to their work, our time-series data can be attributed to the different states of working within an office. The results from this analysis will provide us with highly granular and contextual information about the time and location of occupant activities in a building—the foundation upon which the following proposed work will build.

Inferring occupant activity states from plug load sensor data.

Inferring occupant social/organizational networks

The structure of organizations and the relationships among the people within them are important characteristics of organizational success. Typical organizational charts describe relationships according to team and hierarchical structure, but it is well known in the organizational behavior literature that the true structure of organizations is much more complex, subtle, and difficult to measure accurately. We theorize that various aspects of occupant dynamics influence how two occupants are related to one another, including spatial configuration of the building, the organizational structure of the group in the building, and the social relationships/friendships among the occupants. Given time-series information—inferred from plug load data—about entities that are theorized to have relationships, various models can be utilized to infer the network structure of the entities. In the occupant network, the occupants themselves are the nodes, and the relationships among them are represented in the edge weightings (Figure 2). The influence model (Pan et al. 2012) and the graphical lasso (Friedman et al. 2008) are two models that have been shown to be effective in modeling relationships among entities given time-series data. We propose to investigate the effectiveness of these models in capturing true organizational structures within a building. We have also developed our own occupant inference model, which we call the “interaction model” that we have shown to outperform both the influence model and the interaction model. This mode makes use of our engineering knowledge on building occupant dynamics to infer these ties among occupants. Opportunity for social interaction comes about when occupants are expected to be in the building, but not working at their individual desks. This middle-ground state is likely captured by the medium energy activity state (yellow activity state in Figures 2 and 3). The more often two occupants share this medium state over time, the more likely they are to interact in person.

We have benchmarked the inferred network by capturing ground truth relationship information collected through surveys. By leveraging correlation metrics designed for graphs and testing significance through a permutation test, we found that our interaction model captures a significant amount of the network structure reported through surveys.

Inferring network structure from activity data.

Optimizing layouts for energy and organizational goals

Methods for understanding building occupant activities and network structure within buildings both have inherent value. Learning activities and the corresponding space utilization patterns can enable facilities to more closely match the operation of their systems with the dynamics of occupants. Learning network structure can help organizational managers understand the true ways in which their organizations function. We believe that in addition to these natural upshots, these methods will additionally enable the creation of tools for optimizing the layouts of buildings for our goals of energy efficiency and occupant performance.

As discussed in above, the design of office layouts can have significant impacts on collaboration and productivity. As a result, we can think about suggesting new layouts to promote such outcomes. Having learned the structure of the occupant network, we can attempt to strategically place occupants across the floorplan with strong organizational or social ties in order to promote communication and collaboration. Utilizing the language of space syntax, we can distill layouts into individual spaces and the connections among them. By mapping our learned network onto the space syntax network, we can evaluate how well we promote collaboration.

Any change in the layout of occupants is likely to have significant impacts on the energy performance of the building, as discussed in above. Having developed a method for understanding the activities of individual occupants, we can also evaluate how much different layouts are expected to affect the energy consumption of the building. Beyond evaluating the energy impacts of layout suggestions, we can also optimize layouts for energy-efficiency by clustering occupants with similar schedules. If we can colocate individuals who tend to arrive at the office around the same time, for example, we can leverage the controllability of shared HVAC and lighting services to delay operation until that group’s typical start time. In preliminary analysis, we have found that a simple hierarchical clustering approach based on the Euclidean distance between occupants’ time-series data vectors can produce groups of occupants with similarities in their activities. In Figure 4, we show a dendrogram of this clustering approach on a small office in Berkeley, CA. We see three relatively strong clusters as well as three relative outliers. In this preliminary proof-of-concept work, by grouping the strong clusters into the same zones within the building and simulating the energy performance of the building using EnergyPlus and OpenStudio software tools, we find that this simple clustering approach can save up to 3.3% in total annual energy consumption.

Grouping occupants with hierarchical clustering.

In this research, we aim to refine our methods for optimizing layouts for energy efficiency and combine them with methods that leverage space syntax and network structure to co-optimize for both energy efficiency and occupant performance.

Expected Contributions to Practice

The overarching outcome of this work is the development of a data-driven socio-technical system for facility operations that can create designs and layouts that promote workplace collaboration and productivity while simultaneously reducing energy consumption in commercial buildings.  This framework will serve as a toolset for commercial building management optimization and re-design. Using simple smart plug load sensor devices—or other similar IoT sensors—a facility manager will be able to conduct a “socio-space audit” of their building to understand how their spaces are being used and their organization is functioning in near real-time. Then, a facility manager will be able to utilize our system to explore and implement a set of optimized layouts designed to reduce energy use and encourage collaboration.

Expected Contributions to Theory

From a theoretical perspective, we expect our contributions to center around new methodologies for inferring spatio-social networks in the built environment from sensor data as well as for optimizing commercial building layouts for energy and organizational goals. In the end, this work aims to enable facility managers to simultaneously fine-tune the performance of their buildings and their most important stakeholders—the occupants in their facilities.

Publications

  1. A. J. Sonta and R. K. Jain. (in prep). “Learning human network structure in buildings.”
  2. A. J. Sonta and R. K. Jain. (2019). “Data-driven building layout optimization for energy efficiency,” in International Conference on Applied Energy, (Vasteras, Sweden), Elsevier. [Invited for submission to Applied Energy]
  3. A. J. Sonta and R. K. Jain. (2019). “Building relationships: Using embedded plug load sensors for occupant network inference,” IEEE Embedded Systems Letters. [Invited Paper]

References

  • Bafna, S. (2003). “Space syntax: A brief introduction to its logic and analytical techniques.” Environment and Behavior, SAGE Publications, 35(1), 17–29.
  • Congdon, C., Yan Zhang, Rashid, M., Bafna, S., Warmels, S., Bromberg, J., Bajaj, R., Zimring, C., and Peponis, J. (2007). “Designing Space to Support Knowledge Work.” Environment and Behavior, 39(6), 815–840.
  • Costa, T., Boccignone, G., and Ferraro, M. (2012). “Gaussian mixture model of heart rate variability.” PLoS ONE, (J. Bourdon, ed.), Public Library of Science, 7(5), e37731.
  • D’Oca, S., Hong, T., Langevin, J., D’Oca, S., Hong, T., and Langevin, J. (2018). “The human dimensions of energy use in buildings: A review.” Renewable and Sustainable Energy Reviews, Pergamon, 81, 731–742.
  • Dobbs, J. R., and Hencey, B. M. (2014). “Predictive {HVAC} control using a Markov occupancy model.” 2014 American Control Conference, IEEE.
  • Feng, X., Yan, D., and Hong, T. (2015). “Simulation of occupancy in buildings.” Energy and Buildings, 87, 348–359.
  • Friedman, J., Hastie, T., and Tibshirani, R. (2008). “Sparse inverse covariance estimation with the graphical lasso.” Biostatistics, Oxford University Press, 9(3), 432–441.
  • Hong, T., and Lin, H. (2012). “Occupant Behavior : Impact on Energy Use of Private Offices.” Asim IBSPA Asia Conference.
  • Kabo, F. (2018). “The architecture of network collective intelligence: Correlations between social network structure, spatial layout and prestige outcomes in an office.” Philosophical Transactions of the Royal Society B: Biological Sciences, The Royal Society, 373(1753), 20170238.
  • Kabo, F., Hwang, Y., Levenstein, M., and Owen-Smith, J. (2015). “Shared Paths to the Lab: A Sociospatial Network Analysis of Collaboration.” Environment and Behavior, SAGE PublicationsSage CA: Los Angeles, CA, 47(1), 57–84.
  • Kabo, F. W., Cotton-Nessler, N., Hwang, Y., Levenstein, M. C., and Owen-Smith, J. (2014). “Proximity effects on the dynamics and outcomes of scientific collaborations.” Research Policy, North-Holland, 43(9), 1469–1485.
  • Norford, L. K., Socolow, R. H., Hsieh, E. S., and Spadaro, G. V. (1994). “Two-to-one discrepancy between measured and predicted performance of a ‘low-energy’ office building: insights from a reconciliation based on the DOE-2 model.” Energy and Buildings, 21(2), 121–131.
  • Pan, W., Dong, W., Cebrian, M., Kim, T., Fowler, J. H., and Pentland, A. S. (Sandy). (2012). “Modeling Dynamical Influence in Human Interaction: Using data to make better inferences about influence within social systems.” IEEE Signal Processing Magazine, Institute of Electrical and Electronics Engineers ({IEEE}), 29(2), 77–86.
  • Sailer, K., and McCulloh, I. (2012). “Social networks and spatial configuration—How office layouts drive social interaction.” Social Networks, Elsevier B.V., 34(1), 47–58.
  • Sonta, A. J., Simmons, P. E., and Jain, R. K. (2018). “Understanding building occupant activities at scale: An integrated knowledge-based and data-driven approach.” Advanced Engineering Informatics, Elsevier, 37, 1–13.
  • Waber, B., Magnolfi, J., and Lindsay, G. (2014). “Workspaces that move people.” Harvard Business Review.
  • Yang, Z., Ghahramani, A., and Becerik-Gerber, B. (2016). “Building occupancy diversity and HVAC (heating, ventilation, and air conditioning) system energy efficiency.” Energy, Pergamon, 109, 641–649.
  • Zhao, J., Lasternas, B., Lam, K. P., Yun, R., and Loftness, V. (2014). “Occupant behavior and schedule modeling for building energy simulation through office appliance power consumption data mining.” Energy and Buildings, 82, 341–355.

Original Research Proposal

 

Final Project Report

TR241: Building Relationships: Using Embedded Plug Load Sensors for Occupant Network Inference

TR242: Learning Socio-organizational Network Structure in Buildings with Ambient Sensing Data

Funding Year: 
2020
Stakeholder Categories: 
Owners
Users
Operators/Facility Managers