Towards a Cloud Fog Edge System for Smart Building
This paper tackles privacy, energy, and infrastructure constraints in smart buildings by proposing a decentralized cloud-fog-edge architecture that treats the building itself as a data center. Its core idea couples the KOptim container orchestrator with FIWARE for dynamic SLA management, while running lightweight online ML algorithms such as GHA-PCA and K-means directly on Arduino-class microcontrollers. The humanitarian motivation—bringing autonomous AI to underserved communities—is compelling, although the current evaluation remains largely preliminary.
The paper offers a vision and early prototypes for a cloud-fog-edge smart building system, but its contributions are unevenly substantiated. The architectural proposal combining KOptim and FIWARE is described at a high level, yet the authors admit that KOptim's source code is inaccessible and its mechanics are omitted due to space constraints. The edge-AI experiments port standard online algorithms to ESP32 boards using real sensor traces, but they lack rigorous baselines, convergence analysis, or comparison against established embedded ML frameworks like TensorFlow Lite Micro. While the authors are commendably transparent about these limitations, the current manuscript reads more as a position and progress report than as a fully validated systems paper.
The strongest aspect of the paper is its clear humanitarian framing and use of real-world datasets. The authors ground their motivation in genuine energy and equity concerns, citing current ICT electricity consumption trends, and they validate their edge algorithms on actual traces from the Tour Perret building and CampusIoT LoRaWAN rather than synthetic data. Their explicit acknowledgment of architectural limitations lends intellectual honesty to the work.
The primary weakness is that the paper's central architectural contribution—KOptim coupled with FIWARE—is neither fully explained nor reproducibly evaluated. The authors state that KOptim's workings are omitted "due to space constraints" and that its scheduling source code "remains inaccessible to the public," leaving readers unable to assess the dynamic SLA negotiation claims. On the ML side, the experiments are preliminary: the GHA-PCA C++ and Python implementations yield different results, and the analysis merely lists possible causes without a principled fix. The K-means evaluation on ESP32 reports only message throughput (e.g., processing 1024 messages with batches of 128) rather than clustering accuracy, energy consumption, or memory profiling against comparable microcontroller ML baselines.
The evidence does not consistently support the paper's broader claims about a deployable cloud-fog-edge system. While the GitHub repository provides Arduino code for GHA-PCA and K-means, the FIWARE integration is depicted only in a high-level diagram (Figure 2) with no empirical evaluation of SLA negotiation latency or orchestration behavior under load. Comparisons to related work are fair in acknowledging differences—e.g., versus U-TOE's neural network focus—but serve mostly to delimit scope rather than benchmark performance. The paper does not demonstrate that its online algorithms outperform or even match batch or federated alternatives, nor does it quantify the privacy or energy benefits asserted in the abstract.
Reproduction is partially blocked by the closed-source KOptim framework and incomplete experimental detail. Although some Arduino/ESP32 algorithmic implementations are publicly available on GitHub, the orchestrator-FIWARE coupling cannot be independently validated because "the experiment's source code... remains inaccessible to the public." For the edge ML experiments, hyperparameters such as learning rates, exact convergence thresholds for GHA, and K-means centroid initialization are either missing or buried in repository files without textual description. The authors note ongoing work to set up monitoring for dynamic resources, suggesting that the reported integration is not yet in a reproducible state.
In this article, we present our vision and recent advancements toward creating a decentralized system capable of learning from real-time data within buildings to support sustainable and privacy-preserving smart environments. Our approach promotes the concept of the building itself as the data center, aligning with the principles of edge computing to safeguard confidentiality and reduce reliance on external cloud infrastructure. This is particularly valuable in humanitarian contexts, where data sovereignty, energy efficiency, and infrastructure constraints are critical. We detail a lightweight, "Kubernetes-like" orchestration framework for deploying AI services within such environments and demonstrate our progress in implementing AI algorithms on low-power, cost-effective microcontrollers such as those in the Arduino ecosystem. By enabling in-situ learning directly on sensors or microcontrollers, our work aims to bring intelligent services to resource-limited settings, fostering autonomy, resilience, and sustainable development in vulnerable or underserved communities. The contributions in this article are related, firstly, to our project "Online Machine Learning Algorithms for Embedded Systems" and the evaluation of two new online algorithms. Secondly, we envision a cloud-fog-edge architecture based on the KOptim and FIWARE components, and we propose a methodology for coupling them. Experimental results of the online algorithms are also presented, showcasing real-world traces.
Pick a starting point or write your own. Challenges run in the background, so you can keep reading while the AI investigates.
No challenges yet. Disagree with the review? Ask the AI to revisit a specific claim.