The core of the research is a multi-objective optimization model that weighs four competing goals at once: energy consumption, thermal comfort, visual comfort, and full life cycle cost. Thermal comfort is quantified through the Predicted Percentage of Dissatisfied occupants, or PPD, which is derived from the Fanger PMV index, a heat-balance equation that accounts for metabolic rate, clothing insulation, air temperature, radiant temperature, humidity, and air velocity. Visual comfort is measured by the Daylight Factor, the ratio of indoor to outdoor natural illumination under an overcast sky, while glare is policed by the Unified Glare Rating, which the study treats as a hard constraint capped at 16 rather than a soft objective. Crucially, the model recognizes that these objectives are physically coupled: the window-to-wall ratio and the solar heat gain coefficient jointly determine both how much sunlight enters a room and how much heat rides in with it, creating a nonlinear trade-off that no single-parameter design can resolve.

To capture that coupling, the study builds a predictive engine based on an improved Genetic Algorithm-Back Propagation neural network, or GA-BP. A genetic algorithm scours the space of initial network weights and thresholds to find a favorable starting point, reducing the risk of getting trapped in local minima, and then backpropagation fine-tunes those parameters through error-driven learning. Two refinements set this version apart from the standard recipe. First, the random seed of the Mersenne Twister generator is fixed at 1, so every training run begins from identical initial conditions, making results reproducible and eliminating random fluctuations that would otherwise muddy the analysis of which design variables matter most. Second, instead of a fixed training-test split, the researchers employ a metabolic dataset partitioning method: samples are ordered chronologically, and at each prediction step only the preceding observations train the model while the next observation is reserved exclusively for testing. The window then slides forward one step at a time, so the network is always learning from the freshest data, which improves its adaptability to short-term trends and sudden changes such as heating load swings.

The payoff in prediction accuracy is substantial. The improved GA-BP achieved a coefficient of determination of 0.961, well above the 0.895 of the standard GA-BP, and its root mean square error of 0.174 undercut ARIMA, support vector regression, and even long short-term memory networks. In multi-step forecasting, the mean absolute error rose only from 0.103 at one hour to 0.158 at 24 hours, indicating that the model remains stable over a full-day prediction horizon, exactly what a predictive controller needs to plan ahead.

On top of this predictor sits the neural network model predictive control framework, or NNMPC. Rather than reacting to threshold violations like a conventional thermostat, NNMPC looks into the future: at each sampling instant, the GA-BP model forecasts indoor temperature and illuminance over a prediction horizon, and the controller solves a constrained optimization problem to find the best sequence of control actions, which include air-conditioning supply temperature and lighting power. Only the first action in the sequence is executed; then new sensor measurements arrive, the horizon rolls forward, and the optimization repeats. This receding-horizon mechanism, weighted across energy consumption, thermal deviation, and illuminance deviation, forms a closed loop of state sensing, prediction and optimization, control execution, and feedback correction.

Finding the optimal static design configuration is the job of a parallel hybrid optimizer that fuses NSGA-II, a genetic algorithm strong at global exploration but slow to converge, with MOPSO, a particle swarm method that converges quickly but can stagnate in local optima. In each iteration, both algorithms generate 50 offspring from a population of 100, and fast non-dominated sorting selects the next generation while duplicate Pareto solutions are pruned to preserve diversity. The hybrid’s Inverted Generational Distance, a measure of how close and evenly spread the solutions are relative to the true Pareto front, reached 0.039, more than 40 percent lower than either single algorithm, and it converged after just 118 iterations compared with 156 for NSGA-II and 143 for MOPSO. The computational cost of 263.5 seconds is acceptable because this optimization runs offline; real-time decisions are left to NNMPC.

The case study applied the framework to a five-story middle school teaching building in Tianjin, China, with a gross floor area of 4,532.41 square meters, using typical meteorological year data from the Beijing region of the China Standard Weather Database. Eight design variables were optimized, including exterior wall and roof insulation thicknesses and window parameters. Among the Pareto solutions, the best energy design achieved 60.53 kilowatt-hours per square meter per year, the best thermal comfort solution cut dissatisfaction to 22.85 percent, and the best daylighting solution reached a daylight factor of 3.42 percent, though with glare near its limit. A cost-priority design delivered the largest life-cycle savings of 479,138 yuan but sacrificed daylighting, dropping the daylight factor from 1.85 percent to 1.52 percent, a clear illustration of the trade-off between economics and visual comfort. The comprehensive balanced solution, which performed at a medium-to-optimal level across all four objectives, emerged as the most practical choice for conventional buildings.

The dynamic control layer then proved its worth in simulation and in the field. Under NNMPC, indoor temperature held near 18 degrees Celsius and 300 lux during working hours on a typical day, and on a hot summer day the system kept interiors at 26 degrees even when outdoor temperatures exceeded 30. Compared with traditional threshold-based PI control, the sensor-linked lighting and air-conditioning strategy cut total annual energy consumption from 73.11 to 60.37 kilowatt-hours per square meter per year, saved 52,400 yuan in life-cycle cost, and lifted user satisfaction from 68.2 to 86.3 percent. Field verification confirmed the simulation: indoor temperature deviation shrank from plus or minus 1.3 degrees to plus or minus 0.6 degrees, illumination deviation fell by 55.3 percent, and HVAC and lighting energy use dropped by 14.7 percent and 19.9 percent respectively.

Robustness testing across climates revealed both strengths and limits. In Tianjin, whose climate resembles Beijing’s, energy, comfort, and cost indicators shifted by at most about 1.7 percent. But in Shanghai’s hot, humid summers the energy loss rate reached 8.1 percent, and in Guangzhou, with longer and more intense cooling seasons, it climbed to 14.9 percent, showing that the Beijing-trained model generalizes well to adjacent climates but struggles where thermal and solar conditions diverge sharply. The author suggests that climate-specific transfer learning or online model updating could close that gap, along with small-sample learning techniques for new buildings that lack the historical data on which prediction accuracy depends.

The broader significance is architectural as much as computational. By embedding a fast, reproducible neural predictor inside a rolling-horizon controller, the study demonstrates that the century-old conflict between daylight and heat need not be settled once at the drawing board and then endured forever. Instead, the envelope can be optimized offline for lifetime economics and comfort, while sensors and a predictive brain continuously renegotiate the balance between sunlight, lamps, and air conditioning as weather and occupancy shift hour by hour. For school buildings, where poor lighting has been linked to rising myopia rates and poor thermal conditions to reduced learning efficiency, the 18-point gain in user satisfaction may prove the most consequential number of all.

Subject of Research: Multi-objective optimization and neural network model predictive control of indoor thermal and lighting environments in buildings

Article Title: Multi-objective intelligent optimization and control of indoor thermal and light environment based on NNMPC

Article References: Mei, Y. (2026). Multi-objective intelligent optimization and control of indoor thermal and light environment based on NNMPC. Discover Artificial Intelligence, 6(1), Article 1216. https://doi.org/10.1007/s44163-026-02299-5

Image Credits: AI Generated

DOI: 10.1007/s44163-026-02299-5

Keywords: NNMPC, GA-BP neural network, multi-objective optimization, thermal comfort, daylighting, building energy efficiency, NSGA-II, MOPSO, model predictive control, smart buildings, life-cycle cost, sensor-based control

The research, published in Frontiers in Computer Science, introduces the Architectural Robot Acceptance Model, or ARAM, a threshold-based model that explains how occupants respond to individual adaptations performed by roboticised building elements. The work addresses a growing tension in contemporary architecture and computing. Partitions, ceilings, desks, sofas and even beds can now physically change shape or relocate autonomously, promising gains in comfort, satisfaction and wellbeing. Yet occupants frequently override, avoid or attempt to disable such systems when adaptations feel disruptive, unpredictable or beyond their control. A single unacceptable action, prior human-robot interaction research suggests, can already damage human trust in a robot, making the timing and justification of each autonomous movement a high-stakes design decision.

To build the model, the team re-analyzed interview and observational data from three previous studies of robotic partitions, treating each documented adaptation as a unit of analysis. The three studies spanned markedly different contexts: a semi-immersive cross-reality simulation in eight private homes, a Wizard-of-Oz controlled physical partition in a semi-public office breakout space with 26 participants, and a five-week in-the-wild deployment of a semi-autonomous partition in a shared office with 13 occupants. Across these settings, occupants experienced a combined 247 adaptations, of which 125, or 50.6 percent, were sufficiently documented to enter the qualitative analysis. The re-analysis was conducted through an interpretivist reflexive thematic process involving ten co-authors with expertise in human-building interaction, human-robot interaction, environmental and social psychology, anthropology and human comfort, generating 42 and 36 initial codes from two complementary coding passes that were consolidated into 56 codes, 13 candidate thematic clusters and ultimately three themes.

Those three themes crystallized into three acceptance criteria. First, personal need: occupants had to recognize an unmet need that motivated the adaptation, whether reducing a disturbance such as glare, noise or lost privacy, or improving the quality of their spatial context. Second, adaptation capability: occupants had to expect the robot to actually address that need through its architectural and robotic affordances, an expectation shaped by past experience, colleague recommendations and perceptions of the partition’s size, speed and behavior. Third, social justification: occupants had to consider the adaptation defensible in the eyes of co-located others, weighing whether it disturbed colleagues, violated social norms, made them appear selfish or drew unwanted attention. The pattern that emerged was consistent: occupants tended to accept an adaptation when they recognized a personal need that the architectural robot was capable of addressing in a socially justifiable way.

Depending on how these criteria combined, occupants adopted one of four distinct responses. They endorsed an adaptation, engaging with it as intended, in 56 of the analyzed cases, or 44.8 percent, when all three criteria were met. They refrained from engaging in 38 cases, or 30.4 percent, when the adaptation seemed capable but socially unjustifiable, for instance because moving the partition would darken a colleague’s workspace or disturb an ongoing meeting. They repurposed an adaptation in nine cases, or 7.2 percent, using an otherwise incapable partition in unintended ways, such as redirecting it toward noisy colleagues to send a subtle behavioral message. They rejected an adaptation in 22 cases, or 17.6 percent, either because no personal need was recognized or because the adaptation appeared both incapable and socially indefensible. One participant who rejected an adaptation described feeling threatened as the partition approached; another abandoned a desk that had become claustrophobic when the partition closed in too near.

The conceptual leap of ARAM lies in how it organizes these criteria. Rather than treating personal need, capability and social justification as independent variables, the model positions them as three qualitative thresholds along a single continuum of perceived stress severity, meaning how consequential an unmet or anticipated need becomes for an occupant in a particular situation. Personal need and social justification act as lower thresholds: as perceived stress rises, an unrecognized need becomes recognized, and an unjustifiable adaptation becomes justifiable. Adaptation capability, by contrast, acts as an upper threshold: beyond a certain severity, an adaptation once deemed capable may become insufficient, for example when a partition is too small to block intense glare or too slow to respond in time. The relative positions of these thresholds, compared against the occupant’s current stress level, determine which of the four responses becomes possible.

The model is explicitly situated rather than fixed. Threshold positions vary across occupants, adaptations, needs and contexts, and can shift for the same person over time. An occupant who prioritizes social harmony in an unfamiliar setting may consider an adaptation justifiable only at high stress levels, refraining from it at lower levels even when it addresses a genuine need; in a familiar context, the same person may accept it far earlier, trusting that colleagues will understand its purpose. The authors ground this threshold logic in established psychological theory: the transactional model of stress and coping, which frames stress as a product of subjective appraisal; protection motivation theory, which links responses to perceived efficacy of protective actions; and Granovetter’s threshold model of collective behavior, which defines the point at which perceived benefits of acting outweigh social costs.

ARAM also departs meaningfully from established technology acceptance frameworks such as the Technology Acceptance Model and the Unified Theory of Acceptance and Use of Technology. Those models explain adoption at the level of the overall system through constructs like perceived usefulness, effort expectancy and social influence. ARAM instead operates at the level of the individual, situated adaptation, distinguishing responses that conventional models collapse into a single intention-to-use measure. An architectural robot may be perceived as useful and easy to use in general, yet a specific adaptation may still be judged incapable of addressing a particular need at a particular moment. Conversely, an adaptation may be perceived as capable even when its measurable technical performance is limited, if occupants believe it meaningfully reconfigures their space. The authors propose two testable propositions: that independently assessed combinations of the three appraisal states should systematically correspond to the four responses, and that increasing perceived stress severity should produce the ordered sequence of responses implied by an independently established threshold order.

The practical implications extend to how architectural robots might one day estimate acceptance in real time. The authors suggest that personal need could be proxied through sensing of visual, acoustic, thermal and air-quality conditions, alongside spatial factors such as occupancy density and visual exposure, triangulated with occupant feedback and Personal Comfort Models. Adaptation capability could be estimated through onboarding studies, interaction logs and post-adaptation feedback, with demonstrations helping occupants form accurate expectations of what a robot can and cannot do. Social justification could be informed by occupants’ familiarity with co-located others, perceived norms, hierarchy and sense of responsibility. Where reliable estimates exist, a robot could prioritize adaptations likely to be endorsed, delay those likely to be rejected, request confirmation when occupants may refrain, or offer alternatives when repurposing seems probable. The authors also caution against maximizing predictability at all costs, noting that excessive predictability could erode the curiosity and serendipity that unexpected robotic behavior can inspire.

The researchers are careful to scope their claims. ARAM was developed from robotic partitions operating at the architectural space-plan layer, and its transferability to other robotic, automated, intelligent and smart technologies embedded in the built environment remains to be evaluated, particularly for systems differing in materiality, autonomy or social visibility. The qualitative methodology relied on self-reported experiences subject to recall effects, and the model has not yet been independently validated with data collected specifically to test its propositions. Real-time threshold estimation would also raise technical challenges alongside questions of privacy, consent and the appropriateness of continuous behavioral and physiological sensing. Still, the framework offers designers and researchers a common analytical lens for a future in which walls, ceilings and furniture are no longer static backdrops but active agents whose every movement must earn its place in the shared social fabric of indoor life.

Subject of Research: A threshold model of occupant acceptance of architectural robot adaptations based on qualitative re-analysis of three robotic partition studies

Article Title: ARAM: a threshold model for understanding occupant acceptance of architectural robot adaptations

Article References: Nguyen, A. B. V. D., Christoforou, R., Pigliautile, I., Martins Gnecco, V., Pogladič, A., Arko, S., Pisello, A. L., Schaumann, D., Schweiker, M., & Vande Moere, A. (2026). ARAM: a threshold model for understanding occupant acceptance of architectural robot adaptations. Frontiers in Computer Science, 8, Article 1896307. https://doi.org/10.3389/fcomp.2026.1896307

Image Credits: AI Generated

DOI: 10.3389/fcomp.2026.1896307

Keywords: architectural robots, human-robot interaction, human-building interaction, technology acceptance, adaptive architecture, occupant comfort, robotic partitions, environmental psychology, social justification, perceived stress, smart buildings, ARAM model

The framework, called DMCBO, short for Diffusion-driven Multi-objective Collaborative Optimization for Building Clusters, was developed by Yalan Zheng of Zhengzhou Technical College and Taixiang Yin of North China University of Water Resources and Electric Power. Their central insight is that the joint scheduling of equipment settings and load distribution across an entire cluster of buildings can be treated as a high-dimensional data sample, much like an image. Diffusion models, the same family of generative techniques behind modern text-to-image systems, work by progressively corrupting data with Gaussian noise and then learning to reverse that corruption. The researchers reformulated multi-objective building optimization as exactly this kind of conditional denoising process: starting from random noise, the model strips away noise step by step until a complete, Pareto-optimal collaborative schedule for every piece of equipment in every building emerges.

Technically, the forward diffusion process is a fixed Markov chain that adds tiny amounts of Gaussian noise to real scheduling data over one hundred steps, with the noise coefficient increasing linearly from 0.0001 to 0.02 until the original data becomes indistinguishable from standard Gaussian noise. A neural network is then trained to predict the noise that was added at each step, given the noisy schedule and a conditional encoding vector describing the current situation. Once trained, the model can generate fresh solutions at inference time by sampling pure noise and running the reverse process, refining it over one hundred denoising steps into a schedule tailored to the prevailing weather, occupancy, and grid conditions. Because this generation is a gradient-based learning process rather than an iterative search, it avoids the local optima traps that plague traditional multi-objective evolutionary algorithms and the reward oscillations that destabilize deep reinforcement learning.

What makes the system genuinely cluster-aware is its use of graph neural networks. The researchers model the building cluster as a graph mapped onto a radial distribution microgrid topology, where nodes represent buildings and edges encode physical constraints such as maximum line thermal capacities and transformer limits. A two-layer graph convolutional network passes messages between neighboring buildings, fusing spatial-topological dependencies and network constraints into a single graph embedding. This embedding is concatenated with an adjustable preference weight vector to form the condition that steers the denoising trajectory. In effect, when the model generates a control strategy for one building, it simultaneously perceives the state of its neighbors and the limits of the shared electrical infrastructure, suppressing systemic coordination errors from 0.18 down to 0.06 in the team’s experiments.

The multi-objective aspect is handled through an elegant scalarization trick. Each objective, such as energy consumption, operating cost, or thermal comfort, is normalized to a dimensionless value between zero and one using Min-Max normalization based on historical operational bounds. The normalized objectives are then blended into a single scalar condition via a weight vector whose components sum to one. During training, these weights are sampled from a uniform Dirichlet distribution, exposing the model to the entire space of possible preferences so it learns the distribution of the full Pareto front. At deployment, a decision maker can slide the weight vector like an adjustment knob, instantly generating schedules that emphasize energy saving, comfort, or a balanced compromise, all within the same trained model.

The denoising backbone itself is a U-Net adapted to process one- and two-dimensional building scheduling tensors. Diffusion time steps are embedded through sinusoidal positional encodings, while the graph-derived condition is injected into every residual block via Adaptive Layer Normalization. A bottleneck layer with multi-head self-attention captures long-range dependencies across the entire daily scheduling horizon, and symmetric skip connections preserve high-resolution information during reconstruction. The authors trained the model for 300 epochs on dual NVIDIA Tesla V100 GPUs, and the loss curves dropped from 0.75 to below 0.03 with training and validation curves closely coinciding, indicating no overfitting. As the reverse diffusion step count decreased from 90 to 10 during analysis, the noise-prediction error distribution became sharply concentrated at zero, demonstrating extremely high denoising precision at the end of generation.

Validation was carried out on the open-access Building Data Genome Project 2 dataset, which supplied real hourly weather records, operational schedules, and annual electricity, cooling, and heating load profiles across diverse climate zones. Because that dataset provides metered energy data rather than physical building models, the team used United States Department of Energy Commercial Reference Buildings as templates and calibrated them in EnergyPlus, a high-fidelity thermodynamic simulation platform, to create a digital twin consistent with the real data. All results were benchmarked against a rule-based control strategy with static thermostat setpoints of 25 degrees Celsius in summer and 22 degrees in winter, and every experiment was repeated ten independent times to confirm statistical significance.

The headline numbers are impressive. In high-demand summer conditions, DMCBO achieved an average energy-saving rate of 28.6 percent and cut peak load by 38.5 percent, shaving the afternoon peak from roughly 560 kilowatts down to about 440 kilowatts and saving approximately 142.80 US dollars per day. Cumulative daily power savings exceeded 1,000 kilowatt-hours, compared with 400 kilowatt-hours for a traditional collaborative strategy and just 120 kilowatt-hours without coordination. In winter, peak shaving and valley filling reached about 39 and 42 percent respectively, far above the 22 and 25 percent of conventional methods. When the solution sets were projected into objective space, DMCBO dominated the baselines: it reached a hypervolume indicator of 0.95 within 12,000 function evaluations, versus 0.85 for NSGA-II and 0.70 for MOPSO, while keeping operating costs between 140 and 160 dollars against more than 210 dollars for deep reinforcement learning and 260 to 300 dollars for a GAN-based generative approach.

Robustness and scalability proved equally compelling. Under injected stochastic disturbances mimicking sudden weather changes and occupancy swings, the framework confined power fluctuations to a narrow band of plus or minus 5 kilowatts, whereas the deep reinforcement learning baseline swung between minus 15 and plus 20 kilowatts. Scaling from 5 to 50 buildings, the global energy-saving rate actually rose from 28.5 to 31.2 percent, because larger clusters unlock more load-complementary opportunities that the graph encoder can safely exploit within encoded grid boundaries. Computation time grew to 9.8 seconds at the 50-building scale, but that remains comfortably within the 15-minute control horizon used by real building management systems, and the hypervolume index slipped only slightly to 0.88. Ablation experiments confirmed that both the graph neural network module and the choice of 100 diffusion steps, which yields a 50-millisecond inference latency, were essential to this balance of speed, accuracy, and stability.

The authors are candid about limitations. The framework has so far been validated only in simulation, and EnergyPlus simplifies sub-minute mechanical actuation latencies that real HVAC hardware would introduce; they propose cascading PID loops and predictive safety margins to absorb those delays in practice. Data quality, generalization to residential buildings and extreme climates, the computational cost of scaling to city-sized networks, and the black-box nature of the generative process all remain open challenges, with explainable AI and transfer learning flagged as future work. Still, the architecture is designed to slot into standard building management systems as a supervisory layer, transmitting setpoints over protocols such as BACnet or Modbus. If the simulation gains survive contact with physical buildings, the idea of generating optimal energy schedules from noise, guided by a graph of the grid itself, could become a powerful new tool in the push toward decarbonized cities.

Subject of Research: Conditional diffusion models with graph neural networks for multi-objective energy optimization in building clusters

Article Title: Diffusion model-driven multi-objective collaborative optimization for building energy management using graph neural networks

Article References: Zheng, Y., & Yin, T. (2026). Diffusion model-driven multi-objective collaborative optimization for building energy management using graph neural networks. Discover Artificial Intelligence, 6(1), Article 1383. https://doi.org/10.1007/s44163-026-02140-z

Image Credits: AI Generated

DOI: 10.1007/s44163-026-02140-z

Keywords: diffusion models, graph neural networks, building energy management, multi-objective optimization, EnergyPlus, smart buildings, Pareto optimization, deep learning, energy efficiency, HVAC control, microgrids, generative AI

The technology at the heart of the wallpaper is the moist-electric generator, or MEG. These devices absorb water molecules directly from the surrounding air. Once absorbed, the moisture promotes the dissociation and movement of ions within the material, establishing an ion-concentration gradient across the device. That gradient produces a separation of charge, which in turn establishes a voltage between the two sides of the generator. In practical terms, the wallpaper exploits one of the most ubiquitous and overlooked resources in any occupied building: the constant presence of water vapor released by breathing, cooking, and bathing. Rather than burning fuel or waiting for sunlight, the device draws its energy from the humidity that every home already produces around the clock.

Choi’s team deliberately positioned the technology away from the crowded field of high-energy power generation. The MEGs are not intended to compete with solar arrays, grid connections, or batteries in terms of raw output. Instead, they are designed to serve as dedicated power supplies for low-energy electronics: environmental sensors, wireless communications modules, smart-building interfaces, and the vast constellation of devices that make up the growing “Internet of Things.” These gadgets typically sip milliwatts or less, and wiring them all to mains power or servicing their batteries is one of the practical obstacles slowing the spread of distributed sensing. A wall covering that powers them continuously, without plugging in, could remove that obstacle entirely.

One of the most important insights behind the project is that indoor air, often dismissed as uninteresting from an energy standpoint, is actually a remarkably stable resource. “All previous moist electric-generator devices are for outdoor humidity, because there is a lot of moisture out there that is an excellent energy resource,” Choi said. “The problem is that it generates such small amounts of power, and the outdoor environment is not stable because of extreme sunlight or weather. The good thing about an indoor environment is that it maintains a very constant humidity between 30 and 60 percent, and the occupants’ activities like respiration, cooking, and bathing generate additional moisture.” That constancy matters enormously for a generator that depends on a steady flow of water vapor, because fluctuations in humidity translate directly into fluctuations in output.

Turning that insight into a product, however, required solving a genuine engineering problem. Most MEGs reported previously rely on vertical or horizontal structures with an asymmetric distribution of hygroscopic or ionizable materials. In such designs, one region preferentially absorbs moisture while another promotes desorption or evaporation, creating a sustained moisture and ion-concentration gradient that drives charge separation. The architecture works well at the scale of a laboratory sample. But when researchers attempt to scale it up to wall-sized arrays, two problems emerge: the design uses space inefficiently, and it becomes difficult to maintain directional moisture transport uniformly across a large area. A wallpaper that works in a one-square-centimeter patch is of little use if the same physics falls apart across an entire wall.

Choi’s solution draws on his earlier pioneering work in papertronics, the field of building electronic devices on paper substrates as biodegradable alternatives to traditional circuits. The new MEG architecture resembles a microchip laid out on a circuit board, but rendered in paper and printable materials. Glycerol positioned at the edges of each unit captures moisture from the air, while a raised polyvinylpyrrolidone, or PVP, structure in the center, patterned with a wax layer, controls moisture release and evaporation. This spatial arrangement directs moisture transport from the absorption region toward the evaporation region, maintaining a more controlled gradient for continuous power generation. In effect, each small patch of wallpaper behaves like an engineered micro-device rather than a simple slab of hygroscopic material.

Manufacturability was central to the design philosophy from the start. “It was not easy to integrate three different areas into one paper, but I want it all printable so that we can make it for mass production and at a larger scale,” Choi said. The emphasis on printing is what separates this work from many laboratory energy-harvesting demonstrations that never escape the cleanroom. If the absorption region, the evaporation-control region, and the ion-transport structure can all be deposited by printing processes on a paper substrate, the path to covering large wall areas at low cost becomes far more plausible than it would be for devices requiring delicate assembly. Paper-based electronics also carry potential advantages in sustainability, an area Choi’s group has explored in previous work on biodegradable circuits.

Aesthetics and practicality posed a second challenge that pure physics could not solve. Homeowners, the researchers recognized, would not tolerate a wall covered in visible wires and junctions. The team’s answer was to put all of the wiring for the MEG units on the back of the wallpaper, keeping the visible face clean and ordinary. The researchers then experimented with how the individual generators should be connected electrically, linking them in series and in parallel to determine which configuration produced the most power. Interestingly, both methods produced similar results, a finding that gives the designers useful flexibility in how future arrays are wired depending on the voltage and current requirements of the devices they are meant to supply.

Beyond generating electricity, the technology offers a second, perhaps unexpected benefit: humidity management. Buildings already spend enormous amounts of energy on heating, ventilation, and air-conditioning systems whose job includes removing excess moisture from indoor air. Because the wallpaper’s generators actively absorb and transport water vapor, installing them could help regulate a room’s humidity at the same time as they produce power. “We put a lot of energy into HVAC systems to remove moisture from the air, but if we use this concept, we could reduce or control moisture levels while also generating electricity,” Choi said. In that sense, the material performs double duty, offsetting a portion of the energy a building consumes while improving comfort.

For now, Choi sees the MEG arrays as best suited to powering small devices such as environmental sensors or wireless keyboards, and the team hopes to improve the technology for greater power yields in future iterations. Even at modest output levels, the implications are striking. The Internet of Things is projected to encompass billions of connected devices, many of them installed in buildings where batteries must be replaced and cables run at considerable expense. A power source that is printed onto paper, hidden behind ordinary wallpaper, fed by the humidity people generate simply by living in a space, and capable of simultaneously moderating that humidity, addresses several of those pain points at once. The Binghamton work suggests that the next wave of ambient energy harvesting may not come from the roof or the sky at all, but from the quiet, moisture-laden air between the walls we already live with.

Subject of Research: Indoor moisture-driven energy harvesting using moist-electric generator wallpaper

Article Title: Wallpaper that generates power? Binghamton University researchers have developed it

Article References: Wallpaper that generates power? Binghamton University researchers have developed it. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: moist-electric generators, energy harvesting, wallpaper, indoor humidity, Internet of Things, papertronics, Binghamton University, Advanced Energy Materials, sensors, smart buildings, humidity management, printable electronics

The study, conducted by Neda Hoseini Abforosh, Hasan Abolghasem Gorji, and Naser Derakhshani of Iran University of Medical Sciences, set out to clarify the criteria and standards needed to guide the design, implementation, evaluation, and management of smart hospitals. The researchers followed the well-established scoping review framework developed by Arksey and O’Malley, a methodology designed to map the breadth of literature on a topic and identify key concepts and gaps rather than answer a single narrow clinical question. Their search spanned five major databases: PubMed, Scopus, Web of Science, Embase, and ProQuest, covering publications from 2000 through 2024.

The scale of the initial search illustrates just how rapidly this field has grown. The team identified 3,676 publications over the twenty-four-year window, a testament to the explosion of interest in digital transformation within healthcare. After rigorous screening against eligibility criteria, only 43 papers made the final cut. To be included, studies had to focus specifically on standards and requirements for smart hospital design, implementation, evaluation, or management. That steep attrition rate, from nearly four thousand records to fewer than fifty relevant papers, tells its own story: despite enormous enthusiasm, the literature on formal smart hospital standards remains surprisingly thin and fragmented.

From those 43 papers, the researchers distilled six main themes encompassing 29 sub-themes, a framework they describe as the essential DNA of the smart hospital. The first theme, Smart Technological Infrastructure, contains five sub-themes and covers the foundational digital backbone of any intelligent facility: computing systems, networks, interoperability, and the data platforms on which everything else depends. Without robust and seamlessly connected infrastructure, even the most sophisticated clinical applications cannot function reliably. The emphasis on interoperability reflects a long-standing pain point in health informatics, where electronic health records, health management information systems, and integrated hospital information systems often fail to communicate with one another, creating data silos that undermine the very intelligence a smart hospital is meant to deliver.

The second and largest theme, Smart Services Delivery and Patient Interaction, comprises eight sub-themes and shifts the focus from technology itself to how technology transforms care. This domain captures the ways patients actually experience a smart hospital: telemedicine consultations, remote monitoring, mobile health applications, automated appointment and triage systems, and personalized care pathways driven by data analytics. It also encompasses the clinical side, including AI-assisted diagnostics and decision support tools that augment rather than replace clinicians. The prominence of this theme in the review underscores a crucial point that experts in health services research have long emphasized: a smart hospital is not defined by the quantity of gadgets it owns but by whether those technologies measurably improve the patient journey and clinical outcomes.

The third theme, Operational Management and Administration, includes three sub-themes addressing the managerial machinery behind the scenes. Smart hospitals generate enormous volumes of operational data, from bed occupancy and staffing levels to supply chain movements and equipment utilization. The review highlights standards for using this data to optimize resource allocation, streamline workflows, and support evidence-based administrative decisions. Radio frequency identification technology, for example, enables real-time tracking of assets, medications, and even patients through hospital corridors, reducing losses and errors. Building information modeling, known as BIM, extends this intelligence into the physical plant, allowing facility managers to simulate and manage complex building systems with unprecedented precision.

The fourth theme, Regulatory and Maturity, contains four sub-themes and addresses one of the most underappreciated dimensions of the field: governance. The review points to frameworks such as the Joint Commission of Taiwan’s Smart Healthcare Standard, or JCT-SHS, as examples of formal certification schemes that grade hospitals on their digital maturity. Maturity models allow institutions to assess where they stand on the continuum from conventional to fully intelligent operations and to chart a realistic roadmap for progress. The researchers note that regulatory structures are essential for accountability, ensuring that claims of smartness can be independently verified rather than self-declared. This matters for policymakers and investors alike, who need trustworthy benchmarks before committing resources to large-scale digital transformation projects.

The fifth theme, Safety Infrastructure, spans three sub-themes and confronts the risks that come with connectivity. A hospital saturated with networked devices is a hospital with an expanded attack surface for cyberattacks, and healthcare has become one of the most targeted sectors for ransomware and data breaches. The review emphasizes standards for cybersecurity, data privacy, and patient safety safeguards as non-negotiable components of smart hospital design. Safety in this context extends beyond digital threats to include the reliability of AI-driven clinical tools and the resilience of systems on which life-critical functions depend. The authors’ inclusion of safety as a standalone pillar signals a maturing understanding that intelligence without security is a liability rather than an asset.

The sixth theme, Smart Buildings and Environmental Infrastructure, comprises six sub-themes and grounds the digital hospital in its physical reality. Smart hospitals require intelligent heating, ventilation, and air conditioning systems that respond dynamically to occupancy and infection control needs, energy management systems that reduce costs and carbon footprints, and sensor-rich environments that monitor everything from air quality to asset location. This theme reflects a holistic vision in which the building itself is an active participant in care delivery, not merely a container for technology. The researchers argue that integrating all six themes is essential for a comprehensive, multi-dimensional understanding of smart hospital standards, warning that partial approaches that focus on technology alone will fall short.

Perhaps the most consequential message of the review lies in its framing of smart hospitals as complex health-system transformations rather than simple technology upgrades. The authors conclude that technological advancement is inseparable from organisational, human, infrastructural, regulatory, and managerial capacities. A hospital cannot simply purchase intelligence; it must cultivate the workforce skills, governance structures, cultural readiness, and financial models that allow digital tools to take root. The findings offer practical insights for policymakers, hospital managers, healthcare administrators, and investors seeking informed decision-making in the development, implementation, and assessment of smart hospitals, and the authors suggest that applying these standards could ultimately contribute to the evolution of healthcare systems globally. As hospitals worldwide continue their digital march, this framework provides something the field has conspicuously lacked: a shared map of what smart truly means.

Subject of Research: Standards and indicators for smart hospital design, implementation, evaluation, and management

Article Title: Crafting the DNA of smart hospitals: decoding standards and indicators through a scoping review

Article References: Hoseini Abforosh, N., Gorji, H. A., & Derakhshani, N. (2026). Crafting the DNA of smart hospitals: decoding standards and indicators through a scoping review. BMC Health Services Research. https://doi.org/10.1186/s12913-026-15747-w

Image Credits: AI Generated

DOI: 10.1186/s12913-026-15747-w

Keywords: smart hospitals, scoping review, health informatics, digital transformation, interoperability, artificial intelligence, Internet of Things, healthcare standards, hospital management, cybersecurity, smart buildings, telemedicine

The appeal of the approach lies in what it leaves out. Deep reinforcement learning methods such as Deep Q-Networks have demonstrated impressive results in building control, learning nonlinear strategies directly from sensor data without an explicit model of the building’s thermodynamics. Yet their adoption in real building energy management systems has been hampered by practical friction: training instability, high computational cost, and hardware demands that exceed what typical controllers installed in commercial buildings can provide. A tabular learner, by contrast, stores its knowledge as an explicit table mapping situations to action values, which makes it transparent, compact, and fast. The Tanta team’s best-performing controller made decisions in about 45 microseconds per zone and required a Q-table of only 1.6 megabytes, figures that place the method squarely within the reach of resource-constrained building automation hardware.

SARSA, which stands for State-Action-Reward-State-Action, belongs to the family of on-policy temporal-difference learning algorithms. Unlike its off-policy cousin Q-learning, which learns the value of the greedy action regardless of what the agent actually does, SARSA evaluates the policy it is currently following, updating its action values based on the action it actually takes next. This distinction matters in a building context, where exploration must be gentle and where an overly aggressive update can translate into uncomfortable temperature swings for occupants. The on-policy character of SARSA tends to produce more conservative, stable learning, which the authors identified as a key reason for choosing it over deep alternatives for a safety-sensitive application like climate control.

The central technical challenge was inter-zone coupling. In an open-plan office, warm air drifting from a sunlit zone near the facade alters the cooling load of the interior zone beside it, so a controller that observes only its own local temperature is fundamentally flying blind. The researchers addressed this with a coupling-aware observation design: each zone agent receives, in addition to its local conditions, an aggregate summary of the temperatures in neighboring zones. This lightweight information sharing allows each tabular learner to compensate for the heat exchanged across the invisible boundaries of the open floor, without requiring a centralized controller that would scale poorly as the number of zones grows. The result is a decentralized architecture in which simple agents, each with a modest memory footprint, collectively manage a thermally entangled space.

Training took place in a simulated six-zone office environment driven by TMY3 typical meteorological year weather data from the EnergyPlus weather library, ensuring that the agents experienced a realistic annual profile of outdoor temperatures and solar conditions. Occupancy was modeled with stochastic hybrid-work schedules, reflecting the irregular attendance patterns that have become common in post-pandemic offices and that routinely defeat fixed timetables. The reward function was deliberately multi-objective, balancing thermal comfort against energy consumption and penalizing excessive equipment switching, since frequent ON/OFF cycling of HVAC components accelerates wear and shortens equipment life. The best-trained policy settled at an average of 4.1 switching actions per zone per day, a level the authors suggest is compatible with practical equipment longevity.

The headline result, a 39.39 percent reduction in annual HVAC energy use relative to a clearly defined schedule-based baseline, comes with an important caveat of scientific honesty: the comparison target is a conventional scheduled controller, not a state-of-the-art model predictive control system. Even so, the robustness of the finding is strengthened by replication. Across ten independent training runs, the final energy savings converged to 38.0 plus or minus 0.9 percent, indicating that the learning process is reliable rather than the product of a lucky random seed. The authors also conducted a sensitivity analysis over the reward weights, mapping out the trade-off curve between comfort and energy and showing how the balance can be tuned to a building operator’s priorities.

Thermal comfort was assessed against established standards for human occupancy, and the trained controller held violations to 2.85 percent of occupied hours, a figure that suggests the energy savings did not come at the expense of shivering or sweltering occupants. This comfort-energy balance is the perennial tension of building automation: the cheapest way to save energy is simply to switch everything off, but the value of a learning controller lies in discovering subtler strategies, such as pre-conditioning zones ahead of occupancy, exploiting thermal inertia, and modulating output in response to predicted heat flows between zones. The coupling-aware observations appear to be what enables these strategies, since agents that understand their neighbors’ states can coordinate implicitly through the physics of the shared air volume.

The study situates itself within a rapidly expanding literature on reinforcement learning for building energy management, a field that has produced deep learning controllers demonstrated in real buildings and multi-agent frameworks for allocating energy across zones. What distinguishes this contribution is its deliberate minimalism. Where much of the field has pursued ever-larger neural function approximators, the Tanta researchers pursued the opposite direction, asking how much intelligence can be extracted from the simplest possible learner if the state representation is designed thoughtfully. Their answer is that a tabular method, often dismissed as incapable of handling high-dimensional control problems, becomes competitive when the problem is decomposed sensibly and the agents are given just enough information about coupling to act intelligently.

The practical implications are significant for the building industry, where retrofitting advanced control into existing stock is often blocked by cost and complexity. A controller whose entire learned knowledge fits in 1.6 megabytes and whose decision latency is measured in microseconds can run on the embedded processors already present in many building energy management systems, without cloud connectivity or GPU acceleration. That said, the authors are explicit that their results are simulation-based, and embedded field validation remains future work. The gap between simulated and real buildings, with unmodeled sensor noise, actuator degradation, and occupant behavior, is where many promising control algorithms have stumbled, and the transition from the EnergyPlus simulation to a physical open-plan office will be the true test.

Still, the study offers a compelling counterpoint to the assumption that progress in building intelligence must ride on ever-deeper networks. As buildings account for a substantial share of global energy demand and HVAC represents one of the largest end uses in commercial spaces, even fractional improvements compound into meaningful climate impact. If a decades-old, on-policy learning rule, given a clever observation design and a well-balanced reward, can deliver nearly 40 percent savings with hardware that costs almost nothing extra, the barrier to adoption drops dramatically. The work suggests that for many practical control problems, the smartest move may not be a bigger model, but a better-posed question, and it hands building engineers a lightweight, stable, and interpretable tool that could bring learning-based control out of the laboratory and into the ordinary office floor.

Subject of Research: On-policy reinforcement learning with SARSA for energy-efficient multi-zone HVAC control in open-plan office buildings

Article Title: Enhancing energy management in multi-zone buildings using the on-policy reinforcement learning algorithm SARSA

Article References: Attia, M. A., Abdelaal, M. A., & Sallam, E. A. (2026). Enhancing energy management in multi-zone buildings using the on-policy reinforcement learning algorithm SARSA. Neural Computing and Applications, 38(17), Article 717. https://doi.org/10.1007/s00521-026-12445-w

Image Credits: AI Generated

DOI: 10.1007/s00521-026-12445-w

Keywords: SARSA, reinforcement learning, HVAC control, building energy management, open-plan offices, multi-zone buildings, thermal comfort, energy optimization, tabular learning, smart buildings, EnergyPlus simulation, decentralized control

To understand why this result matters, it helps to distinguish supercapacitors from the lithium-ion batteries that dominate consumer electronics. Batteries store large amounts of energy through slow chemical reactions, which is why they take hours to charge and degrade over hundreds or thousands of cycles. Supercapacitors, by contrast, store energy electrostatically, in the electric field that forms at the interface between an electrode and an electrolyte. They hold relatively small amounts of energy compared with batteries, but they take that energy in and release it extremely rapidly, and in some designs they can survive millions of charge-discharge cycles without significant degradation. That combination of speed and durability makes them attractive for applications where power is needed in short, repeated bursts rather than as a long-duration reserve. For a building fitted with solar panels, supercapacitors embedded in the structure could absorb surges of renewable electricity as they arrive and discharge them on demand, smoothing the mismatch between when energy is generated and when it is actually used.

The central challenge in turning cement into a supercapacitor is that ordinary concrete is a poor electrical conductor. Cement hydrates into a porous, mineral-rich matrix that is excellent at bearing compressive loads but terrible at shuttling electrons and ions. The research team’s solution was to build conductivity directly into the material. They mixed carbon nanotubes, carbon black, and cement together to form a printable electrode ink. Carbon nanotubes are cylindrical structures of carbon atoms just nanometers in diameter, renowned for their exceptional electrical conductivity and mechanical strength. Carbon black, a much cheaper amorphous form of carbon, provides additional conductive pathways and helps form a percolating network throughout the composite. Blended into wet cement, these carbon additives transform the paste from an insulator into an electrode material, while the cement itself continues to cure and harden much as it would in any construction application.

Geometry proved to be just as important as chemistry. Using a 3D printer, the team deposited the electrode ink onto a small concrete slab in a pattern resembling interlocked fingers, a configuration known in electrochemistry as an interdigitated electrode design. In this arrangement, two sets of parallel electrode fingers alternate with one another, like the interlaced fingers of two hands held together without touching. The significance of this geometry lies in the distance that charged ions must travel. In a conventional two-plate capacitor, ions in the electrolyte have to migrate across the entire gap separating the electrodes, which slows the device down and wastes energy as internal resistance. With interdigitated electrodes, every finger of one polarity sits immediately adjacent to fingers of the opposite polarity, so ions move only short distances laterally. As the researchers note, this shortened travel distance made the overall supercapacitor more efficient than previous iterations of cement-based energy storage, which had struggled with exactly this ionic transport bottleneck.

A remarkable feature of the design is that the electrolyte is not added as a separate liquid component but arises from the cement itself. As the cement within the slab hydrated, the chemical reaction that gives concrete its strength, its pores filled with water and dissolved ions that could easily travel between the electrodes. In other words, the very process that turns printable ink into hardened concrete simultaneously creates the ionic medium the supercapacitor needs to function. This monolithic approach, in which electrode, electrolyte, and structural support are one continuous piece of material, is what the study’s title refers to in describing monolithic 3D-printed interdigitated cement-based supercapacitors for structural energy storage. It eliminates the interfaces and packaging that complicate conventional devices and means the energy-storage function can, in principle, be printed wherever a builder wants it within a larger concrete element.

Of course, an energy-storage device embedded in a building is useless if it compromises the building’s safety. Concrete in structural applications must meet strict compressive strength standards, and any multifunctional material has to prove it can carry loads without failing. The researchers therefore subjected their cement supercapacitor to mechanical testing and found that its compressive strength was comparable to that of commercial concrete used in slabs and stairs. This is the result that elevates the work from a laboratory curiosity toward practical relevance. Previous attempts at cement-based energy storage often faced a trade-off, in which adding conductive carbon or porosity weakened the material. Here, the team achieved a device that is simultaneously an electrode, an electrolyte-filled capacitor, and a structural material meeting the expectations placed on ordinary construction concrete.

The researchers also demonstrated that their devices do real electrical work. Three supercapacitors printed on the same slab and wired together successfully powered a small array of LEDs, a modest demonstration on its own but an important proof of concept showing that multiple devices fabricated from a single concrete element can be connected into a functioning energy system. Looking ahead, the team envisions these supercapacitors powering everything from emergency lighting to self-powered sensors. That second application deserves particular attention. Modern buildings increasingly depend on distributed sensor networks that monitor temperature, occupancy, air quality, and structural health, and wiring or periodically replacing batteries for thousands of such sensors is costly and cumbersome. Sensors embedded in or on concrete that draw their operating power from the concrete itself would remove that maintenance burden entirely, enabling dense, self-sustaining monitoring of the built environment.

Like any emerging technology, the cement supercapacitor has identified limits, and the team was candid about one of them. The devices operated stably under moderate heating and cooling, but at around zero degrees Fahrenheit, or minus 18 degrees Celsius, their performance started to wane. This matters because the electrolyte in the system is water-based, arising from the hydration of the cement, and water loses its ionic mobility as it approaches freezing. In cold climates, where buildings must perform year-round, a supercapacitor that fades in deep winter would be a serious limitation. The researchers state that future research will focus on fortifying the supercapacitors against cold-weather conditions, an engineering challenge that will likely involve strategies to keep the pore solution ionically active at low temperatures. It is a reminder that multifunctional materials must satisfy the demands of every function they take on, including environmental resilience across the full range of conditions a building experiences.

The broader implications reach toward what researchers call smart buildings, structures that generate, store, and manage energy within their own fabric. As Jing Zhong explains, if building materials could not only support structures but also store energy, sense their surroundings, and even interact with people, buildings would become more than passive shelters. They could become truly smart environments. The study, published in ACS Nano under the title Monolithic 3D-Printed Interdigitated Cement-Based Supercapacitors for Structural Energy Storage, was supported by funding from the Guangdong Hailong Construction Technology Company Limited, a subsidiary of China State Construction International Holdings Limited, an indication that the construction industry itself sees commercial potential in structural energy storage. Considerable work remains before supercapacitor walls appear on job sites, including scaling from small slabs to full structural elements, integrating renewable generation and power electronics, and solving the cold-weather performance gap. But the foundational demonstration is now on the table: a cement-based supercapacitor that stores energy efficiently, prints into interdigitated geometries, powers LEDs, and matches commercial concrete in compressive strength. The walls of future buildings may not merely shelter their occupants. They may quietly hold the charge that keeps those buildings alive.

Subject of Research: 3D-printed cement-based supercapacitors for structural energy storage in smart buildings

Article Title: Cement-based supercapacitors could power next-generation ‘smart’ buildings

Article References: Cement-based supercapacitors could power next-generation ‘smart’ buildings. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: cement supercapacitor, structural energy storage, 3D printing, carbon nanotubes, carbon black, concrete, smart buildings, renewable energy, energy storage, interdigitated electrodes, ACS Nano, self-powered sensors

Researchers Jessica Al Achy and Abdallah Makhoul of the CNRS institut FEMTO-ST at Université Marie et Louis Pasteur in France, together with Hassan Harb of the American University of the Middle East in Kuwait, have introduced a framework called SWIFT-KD, short for Sliding Window Intelligent Federated Transformer Learning with Knowledge Distillation. The system is designed to train a powerful transformer-based forecasting model across large fleets of buildings while keeping every building’s energy records on its own premises. It also tackles a problem that has quietly hampered federated learning in the real world: the sheer cost of moving enormous neural network updates between resource-constrained edge devices and a central coordination server.

The core architectural idea behind SWIFT-KD is hierarchical. Transformers, the family of models that revolutionized natural language processing and have since swept through time-series forecasting, excel at capturing long-range dependencies, but their attention mechanisms grow computationally expensive as input sequences lengthen. For a building whose hourly energy consumption spans months, feeding the entire history into a monolithic transformer is often impractical on the modest hardware installed at the edge. The researchers instead decompose long energy sequences into overlapping segments using a sliding window scheme, processing these segments with a hierarchical arrangement of transformer layers. Lower layers capture fine-grained local patterns, such as the daily rhythm of occupancy and equipment use, while higher layers aggregate segment-level representations into long-term dependencies, such as seasonal drifts in heating demand. The overlapping windows ensure that no temporal boundary severs a genuine pattern, preserving continuity across the decomposed sequence while keeping the per-device computation manageable.

The second innovation addresses communication, the bottleneck that federated learning pioneer McMahan and colleagues identified as far back as their foundational 2017 work on federated averaging. In conventional federated learning, each participating device trains locally and then transmits its full set of model weights, which can number in the millions of parameters, to a server for aggregation. On smart meters and building controllers with limited bandwidth and intermittent connectivity, this exchange becomes prohibitive when repeated over many training rounds. SWIFT-KD replaces weight transmission with federated knowledge distillation. Rather than shipping its parameters, each building sends soft predictions, the model’s output distributions on shared or public proxy data, which act as a compressed distillation of what the local model has learned. The server aggregates these distilled signals into a global model, and the global model is then distilled back to the edges. According to the study, this substitution achieves a 300-fold reduction in communication volume, a figure that transforms the feasibility of large-scale federated deployment on real building hardware.

To test the framework, the team turned to the ASHRAE Great Energy Predictor III dataset, a widely used public benchmark hosted on Kaggle that contains hourly meter readings from more than 1,000 buildings of diverse types and climates. The evaluation focused on 100 heterogeneous buildings, deliberately chosen to stress the statistical challenge that federated learning researchers call non-IID data. Energy consumption patterns differ wildly between an office tower in one climate zone and a warehouse in another, and models trained under federated averaging frequently struggle when local data distributions diverge this sharply. Heterogeneity of this kind is precisely the condition that most often degrades federated systems in practice, making it a demanding proving ground for any new approach.

The results were striking. SWIFT-KD achieved a coefficient of determination, or R-squared, of 0.9708, a root mean squared error of 92.24 kWh, and a mean absolute error of 44.58 kWh. For context, an R-squared approaching unity indicates that the model explains nearly all of the variance in building energy consumption. More remarkable is the comparison against alternatives: the federated framework outperformed standard federated averaging by 21 percent and exceeded even centralized training, in which all data would be pooled in one place, by 13.3 percent. That last figure deserves emphasis, because the conventional wisdom has long held that federated methods necessarily pay an accuracy tax for the privilege of preserving privacy. Here, the distributed approach did not merely match the centralized baseline; it beat it, suggesting that the hierarchical structure and distillation process may itself act as a useful regularizer when data is heterogeneous.

Equally important for practical deployment is the framework’s communication efficiency over the training lifecycle. SWIFT-KD reached its peak predictive performance within just two communication rounds and maintained stable accuracy throughout the remaining rounds. Convergence this rapid compounds the benefits of the 300-fold compression per round: the total network traffic required to train a production-quality model collapses to a small fraction of what weight-based federated averaging would demand. For building operators weighing whether edge intelligence is worth the operational complexity, this combination of fast convergence and lightweight exchanges substantially lowers the barrier. The framework’s design also sidesteps the need for specialized aggregation infrastructure, since distilled predictions are far smaller and easier to combine than full model snapshots.

The broader significance extends beyond the building sector. The study sits at the intersection of three active research currents: the adoption of transformer architectures for time-series forecasting, the maturation of federated learning as a privacy-preserving training paradigm, and the growing use of knowledge distillation not just to shrink models for deployment but to compress the learning process itself. Surveys of federated distillation have catalogued long-standing challenges, including how to generate or select proxy data for distillation and how to prevent the distilled signal from leaking information about local datasets. By demonstrating that distilled federated learning can decisively outperform both weight-based federated learning and centralized training on a realistic, heterogeneous benchmark, the authors provide evidence that these challenges are tractable at scale.

The implications for smart buildings and the energy transition are immediate. City-scale energy management, demand-response programs, and grid decarbonization all depend on forecasts that respect both accuracy and privacy. A framework that lets hundreds of buildings collaboratively learn a shared forecasting model, each contributing knowledge without disclosing consumption records that could reveal occupancy patterns or business activities, aligns directly with regulatory requirements and public expectations. The researchers have made their source code, hyperparameter configurations, and exact data partitions publicly available on GitHub, and the ASHRAE dataset itself is open, which should allow other teams to verify, extend, and adapt the approach. Whether SWIFT-KD or its descendants become the standard for privacy-preserving energy analytics in commercial buildings, the study makes a compelling case that the trade-off between data privacy and model performance, long treated as inevitable, can be engineered away. In a field where a single percentage point of forecasting accuracy can translate into meaningful energy savings across a real estate portfolio, a method that simultaneously improves accuracy, slashes communication costs by two orders of magnitude, and eliminates the need to centralize sensitive data is likely to draw sustained attention from both researchers and the industry it aims to serve.

Subject of Research: Privacy-preserving federated transformer learning with knowledge distillation for building energy prediction

Article Title: SWIFT-KD: Sliding Window Intelligent Federated Transformer Learning with Knowledge Distillation for Building Energy Prediction

Article References: Al Achy, J., Harb, H., & Makhoul, A. (2026). SWIFT-KD: Sliding Window Intelligent Federated Transformer Learning with Knowledge Distillation for Building Energy Prediction. Machine Learning, 115(9), Article 216. https://doi.org/10.1007/s10994-026-07153-4

Image Credits: AI Generated

DOI: 10.1007/s10994-026-07153-4

Keywords: federated learning, knowledge distillation, transformer models, energy forecasting, smart buildings, deep learning, privacy-preserving machine learning, time series prediction, edge computing, ASHRAE dataset, communication efficiency, sustainability