The central innovation is not a single machine-learning model but an “intelligent service hub” that coordinates several task-specific algorithms. Load forecasting used a long short-term memory network with two recurrent layers, while a deep neural network handled binary fault classification and a random forest estimated whether equipment required maintenance. A convolutional neural network extracted visual features, natural-language processing transformed maintenance records into numerical representations, and a weighted late-fusion module combined the results. The authors assigned 50% of the fusion weight to sensor and time-series features, 25% to text and 25% to images or video, reflecting the continuous availability of electrical measurements and the contextual value of reports and visual inspections.

Information moves through the framework in a closed loop. Smart meters, Internet of Things sensors, cameras, drones, operational logs and inspection records are collected, cleaned and synchronized before being sent to specialized processing modules. The service hub then combines those inputs with a knowledge base containing grid topology, operating rules, maintenance records, constraints and safety limits. Its outputs can include load forecasts, fault classifications, equipment-failure probabilities, maintenance schedules, power-rerouting recommendations, renewable-energy dispatch, demand-response adjustments and cybersecurity alerts. Results from the resulting actions are fed back into the system so that predictions and control decisions can be revised as operating conditions change.

The experimental dataset contained 20,000 time-ordered samples assembled from public smart-grid resources. It included electrical measurements, environmental variables, renewable-generation information, maintenance records and fault labels, with text and visual inputs included for equipment monitoring. Because the measurements were time dependent, the researchers used the earliest 80% for training, the next 10% for validation and hyperparameter tuning, and the final 10% as an independent test set. They also conducted walk-forward validation with five sequential folds to assess performance under a more operationally realistic sequence of training on earlier observations and testing on later ones. Ten independent runs were reported with means, standard deviations and 95% confidence intervals.

For fault detection, the deep neural network produced probability scores for faulty and non-faulty states, which were assessed across decision thresholds using a receiver operating characteristic curve. The resulting ROC-AUC was 0.970 ± 0.012, with a 95% confidence interval from 0.961 to 0.979. The predictive-maintenance random forest combined historical maintenance data with equipment age, temperature, vibration, loading conditions and previous faults. On the 2,000-sample test set, it correctly classified 1,348 normal cases and 562 maintenance-required cases, while misclassifying 52 normal cases and missing 38 maintenance-required cases. Its precision for the maintenance-required class was 91.53%, recall was 93.67% and the F1-score was 92.59%.

The same predictive system attained 95.5% overall accuracy and 94.98% balanced accuracy, with a Matthews correlation coefficient of 0.894. The authors also evaluated a Weibull reliability model to represent failures associated with early life, random events or aging, using available component-failure histories where possible. Its initial scale and shape parameters were based on engineering assumptions when records were unavailable, and the researchers said they examined parameter sensitivity. In the operational configuration, components with a predicted failure probability above a validation-selected threshold of 0.60 were marked as requiring maintenance.

p>The load-forecasting module used 24-step sliding windows to predict the next demand or equipment-risk state. Across the reported repeated experiments, it achieved a MAPE of 4.8% ± 0.5%, with a 95% confidence interval of 4.4% to 5.2%. In an example daily profile, demand ranged from about 350 MW at night to 450 MW at midday. In another comparison, consumption ranged from 280 to 320 MW while generation ranged from 300 to 340 MW, illustrating the mismatch the optimization system was intended to manage by adjusting conventional generation, renewable use, storage and flexible demand. The paper describes this as a benchmark evaluation and does not report deployment in an operating utility.

For energy optimization, the framework worked over a 24-step horizon, maintained a 10% reserve margin and allowed demand-response adjustments of up to 15% of non-critical controllable load. The experimental baseline included conventional generation scheduling and reactive adjustment, while the optimized condition jointly coordinated load forecasts, renewable-generation predictions, storage, power allocation and demand response. Under identical test-horizon conditions, total modeled operating cost fell from $12,000 to $8,400. The largest reported saving was in conventional generation, which declined from $9,200 to $6,450 as renewable generation was prioritized and dispatch was adjusted; smaller reductions appeared in grid losses, renewable curtailment, peak demand and storage-control costs.

An ablation analysis indicated that the unified design mattered more than any isolated component. Removing sensor fusion produced the largest degradation among the component-removal tests, lowering fault-detection AUC from 0.970 to 0.931, predictive-maintenance accuracy from 95.5% to 91.6%, load-forecasting MAPE from 4.8% to 6.7%, and cost reduction from 30.0% to 23.5%. Removing energy optimization lowered the modeled savings to 16.2%, while a sensor-only model reached an AUC of 0.926 and 22.7% cost reduction. A broader ablation table reported an even larger cost reduction of 18.6% without multimodal fusion and 21.8% without the service hub, although the paper does not reconcile every numerical difference among its tables and figures.

Human oversight is built into the proposed control hierarchy. Routine, lower-risk actions—such as load balancing, energy allocation and maintenance scheduling—can be automated, whereas severe faults, major voltage deviations or cybersecurity alerts are routed to an operator dashboard for approval. The cybersecurity design contemplates false-data injection, replay attacks, denial of service, unauthorized commands and poisoned training data, using checks such as timestamp verification, role-based access, command authentication, anomaly filtering and cross-modal consistency. These safeguards remain architectural proposals within the evaluated framework. The study itself identifies noisy or incomplete sensor data, renewable intermittency, model generalization to larger grids and extreme conditions, cybersecurity exposure, implementation expense and regulatory standardization as unresolved limitations. Large-scale field testing with live streaming data will therefore be required before the reported benchmark gains can be treated as demonstrated utility-scale benefits.

Subject of Research: Evaluation of an autonomous, operator-controllable AI framework for multimodal smart-grid management

Article Title: An artificial intelligence based framework for smart power grid management using an autonomous and controllable intelligent service hub with multimodal document processing

Article References: Zeng, F., Zheng, H., Li, C., Rong, J., & Nong, Y. (2026). An artificial intelligence based framework for smart power grid management using an autonomous and controllable intelligent service hub with multimodal document processing. Frontiers in Energy Research, 14, 1910556. https://doi.org/10.3389/fenrg.2026.1910556

Image Credits: AI Generated

DOI: 10.3389/fenrg.2026.1910556

Keywords: smart grids, artificial intelligence, multimodal data, fault detection, predictive maintenance, load forecasting, energy optimization, renewable energy, demand response, grid cybersecurity, artificial, intelligence

The review, led by Jingjing Xu and colleagues, applied PRISMA 2020-aligned protocols to systematically analyze 187 papers published between 2016 and 2026, drawn from Scopus, Web of Science, and IEEE Xplore. The authors defined 3D path planning broadly, encompassing any approach in which terrain information, whether represented as digital elevation models, 3D point clouds, or elevation-augmented grid maps, directly informs the planning process through the cost function, the constraint set, the state-space representation, or the perceptual pipeline. This deliberately inclusive definition reflects the reality of the current literature, where fully 3D methods and terrain-aware adaptations of 2D planning exist along a continuum rather than as a sharp dichotomy.

The publication trend tells a striking story of a field in take-off mode. During the initial stage from 2016 to 2019, annual publications remained below seven, with most studies focused on foundational algorithmic exploration. The growth stage from 2020 to 2022 saw annual output exceed ten papers, driven by the maturation of 3D sensing technologies. Then came a clear inflection point in 2023, after which publications increased exponentially, reaching 39 papers in 2024 and 53 in 2025. This surge aligns with two parallel developments: the rapid proliferation of low-cost 3D sensors such as LiDAR and RGB-D cameras, and the substantive expansion of agricultural mechanization into complex terrains where conventional 2D navigation consistently fails.

Why does 3D planning matter so much? The review identifies four core motivations through systematic keyword extraction and clustering analysis. Improving operational efficiency and coverage is the dominant driver, appearing in approximately 55 percent of reviewed papers. In undulating terrain, traditional 2D planning projects 3D topography onto a horizontal plane, causing the actual distance between adjacent paths on slopes to deviate from planned spacing. The result is skips that leave areas untreated, directly translating into yield loss, and overlaps that waste fuel, seed, fertilizer, and time. By incorporating digital elevation models into path generation, 3D planning ensures that adjacent swaths maintain their intended spacing regardless of terrain undulations.

Reducing energy consumption accounts for roughly 22 percent of papers and represents a rapidly growing concern with the emergence of electric agricultural machinery. In hilly and terraced regions, vertical displacement imposes far greater fuel or battery costs than horizontal movement. A geometrically shortest path that repeatedly traverses steep slopes may consume several times more energy than a longer route that follows gentle contour lines. Multiple studies establish energy consumption models that integrate slope, vehicle mass, and dynamic loads, demonstrating that the energy-optimal path frequently diverges from the distance-optimal path. Field-validated results report energy savings of 10 to 22 percent when 3D energy-conscious planning replaces conventional distance-minimizing approaches, a difference that becomes mission-critical for battery-powered vehicles with limited endurance.

Ensuring mechanical stability and safety, addressed in approximately 10.2 percent of papers, tackles the existential risk of rollover accidents on steep terrain. Research on steep-slope vineyards, where inclinations routinely exceed 35 degrees, integrates the robot’s center of mass directly into the path planning cost function, ensuring that every segment of the planned route remains within safe operational bounds. Related work extends this safety consideration to soil conservation, demonstrating that 3D contour-following paths can reduce soil erosion by up to fivefold compared to conventional patterns. These four motivations are cumulative rather than mutually exclusive, and the essence of 3D planning lies in navigating the inherent trade-offs between productivity, sustainability, safety, and autonomy.

The review maps six distinct agricultural scenarios where 3D planning is most needed, each with its own terrain-related challenges. Hilly and mountainous farmland constitutes the most frequently addressed scenario at approximately 28 percent of papers, defined by significant elevation variation and irregular field geometry. Steep-slope vineyards and orchards follow at roughly 18 percent, characterized by slopes exceeding 30 to 35 degrees where rollover risk presents a fundamental safety hazard. Mountain orchards with complex canopy structures account for about 15 percent, where GPS signal degradation and sensor occlusion complicate perception. Uneven paddy fields and wetlands represent approximately 12 percent, where soft, water-saturated ground creates sinkage risk. Terraced and contoured fields, prevalent across East and Southeast Asia, the Mediterranean basin, and parts of South America, account for roughly 10 percent, creating abrupt elevation changes at terrace boundaries that demand specialized transition paths.

On the algorithmic side, the methodological landscape is characterized by a pragmatic fusion of classical optimization, evolutionary computation, and modern perception-driven techniques. Evolutionary and swarm-based metaheuristics constitute the most prevalent approach at approximately 38 percent of papers, with algorithms such as Improved Particle Swarm Optimization, Grey Wolf Optimizer, Whale Optimization Algorithm, and enhanced Genetic Algorithms widely adopted to solve the NP-hard coverage path planning problem under multiple constraints. Graph-search and sampling-based methods account for approximately 28 percent, with the A* algorithm and its variants enhanced with terrain-aware heuristic functions that incorporate slope, elevation differentials, and energy costs. Deep learning and reinforcement learning approaches represent an emerging paradigm at approximately 12 percent, addressing challenges in perception and real-time decision-making that classical optimization struggles to handle. A noteworthy trend is the hierarchical hybrid architecture, which combines global coverage planning using metaheuristics or graph search with local trajectory optimization using deterministic methods or reinforcement learning.

The perceptual foundation of 3D planning rests on a diverse sensing ecosystem. Digital elevation models and satellite remote sensing represent the most widely used data source at approximately 30 percent of terrain-sensing papers, providing pre-existing, large-scale topographic information that can be directly integrated into path planning cost functions. LiDAR-based SLAM constitutes the second major paradigm at approximately 25 percent, providing dense, accurate 3D point clouds particularly valuable in orchards and vineyards where canopy occlusion degrades GPS reliability. RGB-D and stereo vision systems account for approximately 20 percent, deployed for real-time obstacle detection and crop-weed discrimination. Multi-sensor fusion frameworks represent an increasingly dominant trend at approximately 18 percent, integrating LiDAR, IMU, GNSS, and cameras to compensate for individual sensor limitations. Once acquired, 3D terrain data is predominantly processed into 2.5D grid maps, employed in approximately 45 percent of processing methods, where each cell stores elevation, slope, and traversability cost.

Yet the review’s most consequential finding is a pronounced physical integration gap that constitutes the field’s central bottleneck. Kinematic constraints, such as minimum turning radius and nonholonomic limitations, represent the most thoroughly integrated category at approximately 42 percent of the corpus. Dynamic constraints, including velocity, acceleration, and energy limits, appear in roughly 18 percent of reviewed works. Tire-soil interaction, encompassing compaction, slip, and sinkage, is explicitly addressed in fewer than 8 percent of studies. Comprehensive terramechanics models integrating sinkage, slip, drawbar pull, and multi-pass effects are virtually absent from path planning research. This gap is not merely incremental; a path that is geometrically feasible and curvature-continuous may prove physically unexecutable on soft, sloping, or irregular agricultural soils where wheel slip, sinkage, and traction variability dominate vehicle behavior. The persistence of this bottleneck stems from three intersecting challenges: comprehensive terramechanics models require extensive field-specific soil parameterization rarely available at planning scale, embedding full dynamic formulations into real-time path optimization challenges onboard computational capacity, and disciplinary boundaries between path planning, vehicle dynamics, and soil mechanics communities have slowed cross-pollination of methods.

The simulation-to-reality gap compounds this physical bottleneck as a potential credibility threat. Simulation-only validation constitutes the largest category at approximately 53 percent of papers, while controlled field experiments appear in roughly 35 percent and production-scale, long-duration trials in fewer than 10 percent. The documented gap exhibits three consistent patterns. Path tracking accuracy degrades by 30 to 80 percent when moving from simulation to field conditions. Energy savings observed in simulation are frequently attenuated under real terrain variability, with the improved whale optimization algorithm for hilly farmland reporting 22.26 percent energy reduction in DEM-based simulation but related field-validated studies suggesting actual savings closer to 10 to 15 percent once real-time traction losses and operator corrections are accounted for. Coverage rates in irregular fields drop by 2 to 8 percent due to unmodeled boundary irregularities, obstacle occlusions, and sensor blind spots.

Study quality is improving but remains heterogeneous. Applying a custom five-dimension Study Quality Assessment rubric with independent dual scoring, the review found that 41 papers, or 21.9 percent, were classified as High Quality, 104 as Moderate Quality, and 42 as Limited Quality. A clear temporal trend toward improved methodological rigor is evident: the proportion of High Quality studies increased from 11.8 percent in the initial stage to 16.0 percent in the growth stage and further to 25.8 percent in the take-off stage. However, only 34.2 percent of papers report statistical dispersion measures for primary metrics, limiting the reliability of cross-study comparisons, while open-source code or data availability remains the weakest dimension at 12.8 percent, representing a significant barrier to reproducibility.

The geographic distribution of research output reveals a pronounced structural correlation with terrain complexity. East Asia, predominantly China, constitutes the largest research cluster, contributing approximately 60 percent of all papers, with research activities exhibiting clear topographic alignment. Institutions in topographically complex provinces, including Chongqing, Yunnan, Xinjiang, and Shandong, focus on terrain-specific applications such as hilly farmland coverage and mountain orchard navigation. Western and Southern Europe represents the second major cluster at approximately 17 percent, concentrated in countries with steep-slope viticulture and terraced agriculture. The Americas form a third but smaller cluster at approximately 7 percent. This pattern confirms that 3D path planning is fundamentally a terrain-driven research domain, where research intensity is dictated less by general scientific capacity than by the practical urgency imposed by cultivated topography.

The review identifies evaluation fragmentation as a structural impediment to cumulative progress. Metrics such as coverage rate and path length are defined inconsistently across studies, statistical dispersion measures are omitted in over 65 percent of papers, and test conditions vary widely in field geometry, obstacle density, terrain slope, and robot platform specifications. This fragmentation creates a paradoxical situation: while the volume of published research is growing exponentially, the field’s capacity for rigorous comparative assessment has not advanced commensurately. The emergence of open-source benchmarks such as Fields2Cover and Fields2Benchmark signals a positive trajectory, but the field would benefit substantially from a community-endorsed evaluation framework comprising standardized 3D test terrains, a core metric suite with unambiguous operational definitions, and a requirement for publicly reproducible code and parameter settings.

Looking forward, the review identifies five priority research directions calibrated to different stakeholder groups and time horizons. For the research community in the near term, the most urgent priority is the development and adoption of community-standard benchmarks and reporting protocols, running in parallel with the development of physics-informed surrogate models for terramechanics-aware planning that capture dominant tire-soil interaction effects at computational costs compatible with real-time path optimization. For engineers and system developers in the medium term, effort should be directed toward integrated perception-planning architectures with closed-loop field validation, alongside the extension of single-robot planning paradigms to multi-machine collaborative planning for heterogeneous fleets. For policymakers and funding agencies in the long term, investment in shared field-testing infrastructure and translational research programs is essential, together with regulatory pathways and safety standards for autonomous agricultural machinery operating on complex terrain. The trajectory of 3D path planning will ultimately be determined not by algorithmic sophistication alone but by the field’s collective capacity to bridge the identified gaps between simulation and reality, between geometric feasibility and physical executability, and between laboratory capability and economically viable field deployment.

Subject of Research: 3D path planning for autonomous agricultural machinery in complex terrain

Article Title: From flat fields to rugged terrain: A systematic review of 3D path planning for autonomous agricultural machinery

Article References: Xu, J., Wang, R., Xiong, Y., Xie, Y., Zeng, B., & Yang, F. (2026). From flat fields to rugged terrain: A systematic review of 3D path planning for autonomous agricultural machinery. Smart Agricultural Technology, 15, Article 102593. https://doi.org/10.1016/j.atech.2026.102593

Image Credits: AI Generated

DOI: 10.1016/j.atech.2026.102593

Keywords: 3D path planning, autonomous agricultural machinery, terrain-aware navigation, LiDAR SLAM, digital elevation models, rollover prevention, energy optimization, terramechanics, simulation-to-reality gap, Smart Agricultural Technology, flat, fields

The research, led by Keju An and colleagues at the National Laboratory of the Rockies in Golden, Colorado, tackles a stubborn computational bottleneck. Detailed models of adsorption systems—devices where solid sorbent materials grab CO2 molecules from air and then release them when heated and placed under vacuum—are so computationally expensive that simulating a full year of hour-by-hour operation was previously impractical. Without year-long simulations, engineers could not properly explore how machines should behave when weather and grid conditions change around them.

The team’s solution was a physics-informed reduced-order model: a stripped-down but physically faithful mathematical description of a temperature-vacuum swing adsorption (TVSA) cycle, the six-stage process of vacuum extraction, heating, desorption, cooling, pressurization, and adsorption. Crucially, the model retains the phenomena that matter most. It tracks CO2 and water co-adsorption on the sorbent, uses a two-pool formulation in which some CO2 binds quickly to accessible sites while a slower pool represents stubbornly retained molecules, and captures incomplete regeneration—the fact that each cycle never fully clears the sorbent, leaving residual loading that carries over as a memory into the next cycle.

That cycle-to-cycle memory is the heart of the study’s argument. Because the sorbent’s state at the start of each cycle depends on everything that happened before, performance cannot be predicted from any single steady-state snapshot. The loading states act as memory variables that determine the working capacity available in the next cycle, the humidity-dependent adsorption behavior, and the energy required for regeneration. The model tracks six coupled states—fast and slow CO2 pools, free and bound water, gas-phase CO2 concentration, and bed temperature—propagating them forward step by step through the year.

To make the framework fast enough for serious optimization work, the researchers generated roughly 120,000 cycle-resolved simulations using Latin hypercube sampling across environmental conditions, operating strategies, system designs, and initial sorbent states. Each simulated cycle recorded its ambient conditions, control settings, internal sorbent states, and resulting performance. This dataset then trained a compact neural network—a single hidden layer of 39 neurons—that reproduces the physics model’s state transitions at a tiny fraction of the computational cost, creating a surrogate fast enough to embed inside a model predictive control framework.

Before trusting the surrogate, the team benchmarked the physics model against published experimental data from Sendi and colleagues. Across twelve benchmark cases spanning typical temperature and humidity conditions, the reduced-order model reproduced productivity with a mean absolute percentage error of 10.2 percent and electricity demand with an error of 11.4 percent. Deviations grew largest under extreme conditions—high temperature with low humidity, or very cold conditions—reflecting the simplifications inherent in a lumped, zero-dimensional formulation. But the model faithfully captured the direction and relative sensitivity of performance to environmental changes, which matters more for large-scale screening than exact point accuracy.

Sensitivity analysis revealed a starkly asymmetric performance landscape. Regeneration temperature dominated everything: raising it from 70 to 130 degrees Celsius cut annual productivity by roughly 10 tonnes of CO2 per cubic meter of bed while adding 0.7 to 0.8 megawatt-hours of electricity demand per tonne captured—a punishing double penalty from hotter, slower cycles. Humidity was the second most influential factor for energy, with relative humidity climbing from 0.1 to 0.9 adding roughly 0.5 megawatt-hours per tonne, a direct consequence of water co-adsorbing alongside CO2 and demanding extra thermal energy during regeneration.

The real payoff came when the neural-network surrogate was handed to the optimization controller and run against hourly 2025 weather, electricity price, and grid carbon intensity data for two contrasting American locations: humid Atoka, Texas and dry, cooler Madison, Wyoming. A baseline machine running fixed settings completed about 1,094 cycles per year. The adaptive controller pushed that to roughly 1,449 cycles in Atoka and 1,379 in Madison by shifting energy-hungry regeneration steps into hours when electricity was cheap and clean, and exploiting favorable humidity windows for capture.

The results were striking. In Atoka, electricity demand fell 18.8 percent while annual CO2 productivity rose 25.5 percent. In Madison, energy dropped 8.9 percent and productivity climbed 22.4 percent. The most dramatic gains appeared in carbon-adjusted net productivity—capture minus the emissions embedded in the electricity consumed—which nearly doubled at both sites, rising 95.2 percent in Texas and 94.4 percent in Wyoming. Notably, the economics did not always align with the climate math: electricity cost fell slightly in Atoka but rose 3.7 percent in Madison, showing that carbon-aware control sometimes accepts higher operating costs to harvest cleaner power.

The authors are careful about scope. Their model is a control-oriented screening tool, not a plant-specific predictor, and because the surrogate learns from the physics model, its uncertainties propagate into the optimization results—meaning the reported gains are relative, model-based estimates rather than guarantees. Still, the framework offers something the field has lacked: a scalable path from adsorption physics through fast dynamic simulation to real-world control, demonstrating that the biggest untapped efficiency reserve in direct air capture may not be new materials or bigger machines, but smarter timing.

Subject of Research: Dynamic modeling and neural-network-based optimization of solid-sorbent direct air capture systems under varying weather and electricity conditions

Article Title: Cycle-resolved dynamic modeling of atmospheric CO2 separation systems under time-varying ambient and electricity conditions

Article References: An, K., Wald, D., Chen, K., & King, R. (2026). Cycle-resolved dynamic modeling of atmospheric CO2 separation systems under time-varying ambient and electricity conditions. Frontiers in Environmental Engineering, 5, Article 1881554. https://doi.org/10.3389/fenve.2026.1881554

Image Credits: AI Generated

DOI: 10.3389/fenve.2026.1881554

Keywords: direct air capture, carbon dioxide removal, temperature-vacuum swing adsorption, reduced-order model, neural network surrogate, model predictive control, humidity, electricity carbon intensity, energy optimization, solid sorbents, climate engineering, Cycle-resolved

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

The stakes could hardly be higher. Every video stream, industrial sensor reading, autonomous vehicle decision, and healthcare monitoring alert now traverses a layered computing infrastructure in which computation may happen in a hyperscale data center, a regional fog node, or a device at the very edge of the network. Task scheduling, the process of deciding which computing resource will execute which job and when, directly determines latency, throughput, cost, and energy consumption. Load balancing, its close companion, spreads work evenly across available resources to prevent bottlenecks and failures. When either process falters, the consequences ripple outward: service level agreements are violated, users experience lag, servers idle wastefully or overheat, and in safety-critical applications such as connected vehicles or remote surgery, delays can be dangerous. The survey emphasizes that quality of service, resource efficiency, and overall system performance all hinge on how well these two intertwined problems are solved.

What distinguishes this survey from its predecessors is its insistence on a unified, comparative lens. Earlier reviews, the authors note, frequently examined cloud, fog, and edge environments in isolation, obscuring the trade-offs that emerge when workloads migrate between layers. By contrast, the new analysis traces how the optimal scheduling strategy shifts with position in the continuum. In the cloud, where resources are abundant but network latency to end users is significant, meta-heuristic algorithms dominate the literature. These nature-inspired optimizers, including genetic algorithms, particle swarm optimization, ant colony optimization, artificial bee colony methods, grey wolf optimizers, whale optimization algorithms, and cuckoo search, iteratively explore vast search spaces to find near-optimal assignments of tasks to virtual machines. The survey catalogues their underlying mechanisms in detail, from the mutation and crossover operators borrowed from biological evolution to the physics-inspired strategies modeled on phenomena such as simulated annealing and gravitational search.

In fog and edge environments, the picture changes dramatically. Resources are scarce, geographically dispersed, and volatile; network conditions fluctuate; and decisions must often be made in milliseconds. Here, the survey finds, the literature increasingly leans on artificial intelligence and machine learning, particularly reinforcement learning, in which an agent learns a scheduling policy through trial-and-error interaction with its environment, and deep reinforcement learning, which uses deep neural networks to handle enormous state spaces that would overwhelm tabular methods. The reviewed studies span Q-learning approaches for energy-efficient cloud scheduling, deep Q-network variants with improved exploration strategies, multi-agent deep reinforcement learning for distributed fog systems, soft actor-critic methods for task offloading in the internet of vehicles, and meta-reinforcement learning techniques designed to adapt rapidly to unseen edge scenarios. Applications range from smart factories and smart cities to healthcare management systems, where prioritizing critical patients over routine workloads can be a matter of life and death.

The technical anatomy of these learning-based schedulers is one of the survey’s richest contributions. In a typical deep reinforcement learning formulation, the state captures the current load of servers, queue lengths, task characteristics, and network conditions; the action space represents possible task-to-resource assignments or offloading decisions; and the reward function encodes objectives such as minimized makespan, reduced energy consumption, lower latency, or fewer service level agreement violations. Because real deployments rarely optimize a single metric, multi-objective formulations have proliferated, often combining reinforcement learning with meta-heuristics in hybrid schemes. One reviewed study, for example, coupled an artificial bee colony algorithm with reinforcement learning to balance competing objectives in cloud environments, while others integrated fuzzy logic to handle the uncertainty inherent in fog-cloud workflow scheduling. The authors also document the role of federated reinforcement learning, which allows edge nodes to learn collaboratively without centralizing sensitive data, and long short-term memory networks used for proactive scheduling of streaming applications before demand spikes materialize.

Measuring success in this field is itself a subtle science, and the survey devotes substantial effort to standardizing the metrics by which scheduling algorithms should be judged. Its appendix enumerates formal definitions for a comprehensive battery of indicators: makespan, the completion time of the last task in a batch; throughput, the number of tasks completed per unit time; response time and turnaround time; resource utilization; degree of imbalance; energy consumption; monetary cost; migration time; reliability, defined as the ratio of successful executions to total executions; availability; and service level agreement violation, computed from host active time and performance degradation. By assembling these definitions in one place, the authors provide researchers with a common vocabulary, addressing a long-standing problem in which studies claiming improvements often measure incomparable quantities across incomparable testbeds.

That testbed problem receives its own treatment. The survey reviews the simulation ecosystems on which the field depends, from the venerable CloudSim and its extensions such as CloudSim Plus and CloudAnalyst, to iFogSim and FogNetSim++ for fog scenarios, PureEdgeSim for the full cloud-fog-edge spectrum, and GreenCloud, iCanCloud, DCSim, and SCORE for energy-aware data center modeling. The authors classify these tools by environment, features, implementation language, open-source status, and graphical interface, noting that reproducibility and cross-paradigm evaluation remain weak points. Many studies validate their algorithms only within a single simulator tuned to a single layer, making it difficult to know whether a scheduler that excels in a cloud testbed will survive contact with the unpredictable wireless links and device heterogeneity of a real edge deployment.

From this synthesis, the survey distills a set of critical findings and uncomfortable gaps. Performance trade-offs emerge everywhere: meta-heuristics can approach optimal solutions but may converge slowly or demand heavy computation; reinforcement learning agents adapt dynamically but require extensive training and can behave unpredictably under distribution shift. More fundamentally, the authors identify the lack of cross-layer scheduling metrics as a key deficiency, since no standard measure exists to evaluate a decision that spans cloud, fog, and edge simultaneously. They also flag the insufficient handling of dynamic network conditions, pointing out that many algorithms assume stable connectivity even as real-world edge systems contend with fading links, mobility, and congestion. The absence of unified benchmarks spanning the full continuum, and the scarcity of studies addressing security and privacy constraints in scheduling decisions, round out a research agenda that the authors lay out for the community.

The future directions the survey champions read like a blueprint for the next decade of distributed computing. AI-driven adaptive scheduling, in which schedulers continuously learn and re-learn as workloads and network conditions evolve, stands at the forefront, alongside energy-aware optimization that treats power consumption as a first-class objective rather than an afterthought. The authors envision deeper integration of meta-learning so that scheduling agents can generalize across environments, wider adoption of federated approaches that preserve data sovereignty at the edge, and hybrid architectures that combine the global search power of meta-heuristics with the real-time responsiveness of learned policies. As billions of devices join the internet of things and applications from autonomous transport to industrial automation demand guaranteed latency, the survey argues, the systems that manage computation must themselves become intelligent. This work, published as volume 29, article 738 of Cluster Computing, positions itself as a foundational resource for the researchers and practitioners who will build that intelligent infrastructure, offering them a single map of a field that has, until now, been scattered across three separate territories.

Subject of Research: Task scheduling and load balancing techniques across cloud, fog, and edge computing systems

Article Title: Task scheduling and load balancing in cloud, fog, and edge computing systems: a survey

Article References: Zorgati, R., Hamdi, H., & Omri, M. N. (2026). Task scheduling and load balancing in cloud, fog, and edge computing systems: a survey. Cluster Computing, 29(13), Article 738. https://doi.org/10.1007/s10586-026-06513-y

Image Credits: AI Generated

DOI: 10.1007/s10586-026-06513-y

Keywords: task scheduling, load balancing, cloud computing, fog computing, edge computing, reinforcement learning, deep reinforcement learning, metaheuristics, quality of service, energy optimization, internet of things, artificial intelligence