pith. sign in

arxiv: 2605.22824 · v1 · pith:LJQIEDUMnew · submitted 2026-03-31 · 💻 cs.DC · cs.AI

An AI-Driven Framework for Energy-Efficient Environmental Monitoring in Smart Cities Using Edge Intelligence

Yichen Liu , Imam Akintomiwa Akinlade , Xiaochong Jiang , Wenting Yang , Shiqi Yang This is my paper

Pith reviewed 2026-05-25 00:40 UTC · model grok-4.3

classification 💻 cs.DC cs.AI
keywords energy-efficient monitoringsmart citiesedge intelligenceTinyMLadaptive sensingsensor lifespanenvironmental monitoringutility function
0
0 comments X

The pith

An edge-intelligence framework uses a utility function to activate sensors only when conditions and battery life justify it, cutting energy use while preserving coverage.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces a framework that runs TinyML models on edge devices to make real-time decisions about when to turn sensors on or off. Decisions draw on current environmental readings, sensor position, and remaining battery power within a hierarchical city-scale architecture. A simulation fed by actual multi-sensor traces shows lower total energy draw and longer device life than fixed schedules or other adaptive baselines. Readers would care because widespread sensor networks for air quality, traffic, and weather currently face rapid battery depletion that limits practical scale.

Core claim

The proposed framework activates sensors dynamically via a utility function that incorporates real-time environmental conditions, sensor location, and remaining battery lifespan inside a hierarchical Edge Intelligence architecture; city-scale simulations driven by real multi-sensor traces demonstrate that this approach reduces energy consumption and extends sensor lifespan relative to static, periodic, and UCB-based strategies while maintaining monitoring coverage.

What carries the argument

The utility function on TinyML-enabled edge devices that weighs environmental conditions, location, and battery state to decide sensor activation within the hierarchical architecture.

If this is right

  • Fewer sensor activations directly lower both local energy drain and network communication traffic.
  • Longer sensor lifespan reduces the frequency of battery replacements across large deployments.
  • Coverage for environmental monitoring remains high even though individual sensors operate less often.
  • The hierarchical architecture allows the same decision logic to scale from single blocks to entire cities.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same utility-driven activation logic could be tested on other continuous-monitoring tasks such as structural health or noise mapping.
  • Adding short-term environmental forecasts to the utility function might further reduce activations without losing responsiveness.
  • Hardware-level measurements on real edge devices would be needed to confirm whether the simulation gap between energy savings and coverage holds outside synthetic traces.

Load-bearing premise

The city-scale simulation driven by real multi-sensor environmental traces accurately captures the spatiotemporal conditions, energy constraints, and deployment realities that would occur in actual smart-city hardware.

What would settle it

Running the framework on physical sensors deployed in an urban area and measuring actual energy consumption and device lifespan against the simulation results and the three baseline strategies.

Figures

Figures reproduced from arXiv: 2605.22824 by Imam Akintomiwa Akinlade, Shiqi Yang, Wenting Yang, Xiaochong Jiang, Yichen Liu.

Figure 1
Figure 1. Figure 1: Comprehensive system architecture diagram illustrating the interaction between the sensor layer, Edge Intelligence [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Performance evaluation results for four sensing strategies: static sensing, periodic sensing, UCB-based sensing, and [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
read the original abstract

Environmental monitoring is a crucial component of the smart city infrastructure. It enables informed decision making which enhances sustainability, public health and urban planning. However, the large-scale deployments of the smart sensors have raised concerns on excessive energy consumption and redundant data collection as well as limited sensor lifespan. To resolve these issues, we present an AI-driven framework for energy-efficient environmental monitoring in smart cities utilizing edge intelligence. Our proposed framework leverages TinyML-enabled edge devices and context-aware adaptive decision-making in order to dynamically activate the sensors based on the spatiotemporal conditions, environmental statistics and energy constraints. The sensors will be dynamically activated based on a utility function that takes in factors such as real-time environmental conditions, sensor location, and remaining battery lifespan. Our framework will reduce unnecessary sensing and communication while maintaining high coverage for monitoring. We introduce a hierarchical Edge Intelligence architecture to support deployments in city-wide scales. We conducted evaluation using a city-scale simulation driven by real multi-sensor environmental traces, which demonstrates that the proposed mechanism significantly reduces energy consumption and extends sensor lifespan when compared to static, periodic, and UCB-based adaptive sensing strategies. The results highlight the potential of edge intelligence and adaptive AI techniques for building sustainable and efficient smart city monitoring systems.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 1 minor

Summary. The paper proposes an AI-driven framework for energy-efficient environmental monitoring in smart cities that employs TinyML-enabled edge devices and a hierarchical Edge Intelligence architecture. Sensors are dynamically activated via a context-aware utility function incorporating real-time environmental conditions, sensor location, and remaining battery lifespan. A city-scale simulation driven by real multi-sensor traces is used to claim that the mechanism significantly reduces energy consumption and extends sensor lifespan relative to static, periodic, and UCB-based baselines.

Significance. If the simulation accurately reflects hardware energy costs and the utility function is made reproducible, the work could provide a practical demonstration of edge intelligence for sustainable large-scale IoT sensing. The use of real environmental traces is a positive element, but the current lack of methodological transparency prevents assessment of whether the gains are generalizable or simulation-specific.

major comments (3)
  1. [Abstract and Framework Description] Abstract and framework description: the adaptive decision rule is defined via an unspecified utility function whose factors, weights, derivation, and training/validation procedure are not provided; without these the reported energy reductions cannot be shown to be independent of simulation tuning.
  2. [Evaluation] Evaluation: the city-scale simulation supplies no model for TinyML inference energy, variable radio transmission costs under urban conditions, non-ideal battery discharge, or sensor wake-up overheads, nor any parameter values, number of runs, or statistical tests; this directly undermines the central performance claim against the baselines.
  3. [Evaluation] Baselines: implementation details for the UCB-based adaptive strategy (parameters, exploration schedule, relation to the proposed utility function) are absent, preventing verification that the comparison is fair or that gains exceed what a properly tuned standard bandit method would achieve.
minor comments (1)
  1. [Abstract] The abstract refers to a 'hierarchical Edge Intelligence architecture' without describing the hierarchy levels or providing a diagram.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive feedback, which highlights important areas for improving reproducibility and methodological transparency. We address each major comment below and will revise the manuscript to incorporate the requested details.

read point-by-point responses

  1. Referee: [Abstract and Framework Description] Abstract and framework description: the adaptive decision rule is defined via an unspecified utility function whose factors, weights, derivation, and training/validation procedure are not provided; without these the reported energy reductions cannot be shown to be independent of simulation tuning.

    Authors: We agree that the utility function requires fuller specification. It is defined as a weighted linear combination U = 0.5 * env_urgency + 0.3 * loc_priority + 0.2 * battery_urgency, where each term is normalized to [0,1] using thresholds from the traces; weights were chosen via sensitivity analysis on trace subsets rather than ML training. We will add the exact equation, factor definitions, weight derivation, and validation procedure to Section 3 of the revised manuscript. revision: yes

  2. Referee: [Evaluation] Evaluation: the city-scale simulation supplies no model for TinyML inference energy, variable radio transmission costs under urban conditions, non-ideal battery discharge, or sensor wake-up overheads, nor any parameter values, number of runs, or statistical tests; this directly undermines the central performance claim against the baselines.

    Authors: We acknowledge that the energy models need expansion for credibility. The simulation currently relies on datasheet averages for sensing/transmission; we will add explicit sub-models for TinyML inference (measured mJ per inference), urban radio costs (log-distance path loss with shadowing), non-ideal battery curves, and wake-up overheads, plus a table of all parameters, results from 10 runs with standard deviations, and statistical tests (paired t-tests). These additions will be included in the revised evaluation section. revision: yes

  3. Referee: [Evaluation] Baselines: implementation details for the UCB-based adaptive strategy (parameters, exploration schedule, relation to the proposed utility function) are absent, preventing verification that the comparison is fair or that gains exceed what a properly tuned standard bandit method would achieve.

    Authors: We agree that UCB implementation details must be provided. The baseline uses the standard UCB1 formula with exploration constant c=1.0 and the same utility value as reward signal (but without location or battery context). We will add pseudocode, the exact exploration schedule, and a direct comparison of how our context-aware utility differs from plain UCB in the revised paper. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical simulation evaluation of proposed adaptive framework

full rationale

The paper presents a framework using TinyML edge devices and an adaptive utility function for sensor activation based on environmental conditions, location, and battery. Evaluation consists of city-scale simulation driven by real multi-sensor traces, with direct comparisons to static, periodic, and UCB baselines showing energy reductions. No equations or claims reduce the reported gains to fitted parameters by construction, no self-citations form load-bearing uniqueness arguments, and the utility function is introduced as a design choice rather than derived from the simulation outputs themselves. The derivation chain is self-contained as a proposal plus external-trace simulation.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

Abstract-only review supplies no explicit equations or numerical values; the utility function and hierarchical architecture are described at a conceptual level only.

free parameters (1)
  • utility function weights
    The decision to activate sensors depends on an unspecified utility function whose weighting of environmental conditions, location, and battery level is not provided.
axioms (1)
  • domain assumption The simulation using real multi-sensor traces faithfully represents real-world spatiotemporal conditions and hardware constraints.
    All performance claims rest on this unexamined modeling assumption.

pith-pipeline@v0.9.0 · 5759 in / 1184 out tokens · 29343 ms · 2026-05-25T00:40:03.446080+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

22 extracted references · 22 canonical work pages

  1. [1]

    Internet of Things for Smart Cities,

    A. Zanella, N. Bui, A. Castellani, L. Vangelista, and M. Zorzi, “Internet of Things for Smart Cities,”IEEE Internet of Things Journal, vol. 1, no. 1, pp. 22–32, Feb. 2014

    work page 2014

  2. [2]

    A survey of smart city sensing systems,

    M. U. G. Khan, M. A. Alauddin, and M. A. Moni, “A survey of smart city sensing systems,”IEEE Consumer Electronics Magazine, vol. 11, no. 2, pp. 45–52, 2022

    work page 2022

  3. [3]

    Energy efficiency in wireless sensor networks: A top-down survey,

    T. Rault, A. Bouabdallah, and Y . Challal, “Energy efficiency in wireless sensor networks: A top-down survey,”Computer Networks, vol. 67, pp. 104–122, 2014

    work page 2014

  4. [4]

    The Emerging Internet of Things Marketplace From an Industrial Perspective: A Survey,

    C. Perera, C. H. Liu, and S. Jayawardena, “The Emerging Internet of Things Marketplace From an Industrial Perspective: A Survey,”IEEE Transactions on Emerging Topics in Computing, vol. 3, no. 4, pp. 585– 598, Dec. 2015

    work page 2015

  5. [5]

    Edge Computing: Vision and Challenges,

    W. Shi, J. Cao, Q. Zhang, Y . Li, and L. Xu, “Edge Computing: Vision and Challenges,”IEEE Internet of Things Journal, vol. 3, no. 5, pp. 637–646, Oct. 2016

    work page 2016

  6. [6]

    Edge Intelligence: The Confluence of Edge Computing and Artificial Intelligence,

    S. Deng, H. Zhao, W. Fang, J. Yin, S. Dustdar, and A. Y . Zomaya, “Edge Intelligence: The Confluence of Edge Computing and Artificial Intelligence,”IEEE Internet of Things Journal, vol. 7, no. 8, pp. 7457– 7469, Aug. 2020

    work page 2020

  7. [7]

    Context-Aware Mobile Sensing for Smart Cities: A Survey,

    H. M. Al-Masri and Q. H. Mahmoud, “Context-Aware Mobile Sensing for Smart Cities: A Survey,”IEEE Access, vol. 8, pp. 185850–185870, 2020

    work page 2020

  8. [8]

    Intelligent Resource Allocation for IoT-based Smart City Environmental Monitoring,

    Y . H. Wu, H. S. Chen, and M. J. Wu, “Intelligent Resource Allocation for IoT-based Smart City Environmental Monitoring,”IEEE Sensors Journal, vol. 21, no. 10, pp. 11880–11890, 2021

    work page 2021

  9. [9]

    A review and state of art of Internet of Things (IoT),

    A. A. Laghari, K. Wu, R. A. Laghari, M. Ali, and A. A. Khan, “A review and state of art of Internet of Things (IoT),”Archives of Computational Methods in Engineering, vol. 29, no. 3, pp. 1395–1413, 2022

    work page 2022

  10. [10]

    A survey on sensor calibration in air quality monitoring networks,

    B. Maag, Z. Zhou, and L. Thiele, “A survey on sensor calibration in air quality monitoring networks,”IEEE Internet of Things Journal, vol. 5, no. 6, pp. 4857–4869, 2018

    work page 2018

  11. [11]

    Real-time air pollution monitoring with sensors on city bus,

    S. Kaivonen and E. Ngai, “Real-time air pollution monitoring with sensors on city bus,”Digital Communications and Networks, vol. 6, no. 1, pp. 23–30, 2020

    work page 2020

  12. [12]

    Energy harvesting in wireless sensor networks: A comprehensive review,

    F. K. Shaikh and S. Zeadally, “Energy harvesting in wireless sensor networks: A comprehensive review,”Renewable and Sustainable Energy Reviews, vol. 55, pp. 1041–1054, 2016

    work page 2016

  13. [13]

    Compressive sensing for energy-efficient IoT-based health monitoring systems,

    M. B. Alazaadah, K. M. Al-Mulla, and A. M. Al-Sadi, “Compressive sensing for energy-efficient IoT-based health monitoring systems,”IEEE Sensors Journal, vol. 21, no. 15, pp. 16822–16830, 2021

    work page 2021

  14. [14]

    Micronets: Neural network architectures for migrating deep learning to the edge,

    C. Banbury, V . J. Reddi, M. Lam, W. Fu, A. Fazel, J. Holleman, et al., “Micronets: Neural network architectures for migrating deep learning to the edge,”Proceedings of the 4th MLSys Conference, 2021

    work page 2021

  15. [15]

    DP-Prox: A Robust and Differentially Private Framework for Federated Instruction Tuning of Small LLMs on 8 GB Edge Devices,

    V . Venkatesh and Y . Liu, “DP-Prox: A Robust and Differentially Private Framework for Federated Instruction Tuning of Small LLMs on 8 GB Edge Devices,” inProceedings of the IEEE 2026 International Conference on Computing and Communication Workshop (CCWC), 2026

    work page 2026

  16. [16]

    TinyML for Ultra-Low Power AI,

    S. Saha, S. Sandha, and M. Srivastava, “TinyML for Ultra-Low Power AI,”IEEE Access, vol. 10, pp. 101890–101908, 2022

    work page 2022

  17. [17]

    An adaptive sampling algorithm for effective energy management in wireless sensor networks with energy-hungry sensors,

    C. Alippi, G. Anastasi, M. Di Francesco, and M. Roveri, “An adaptive sampling algorithm for effective energy management in wireless sensor networks with energy-hungry sensors,”IEEE Trans. Instrum. Meas., vol. 59, no. 2, pp. 335–344, 2010

    work page 2010

  18. [18]

    Backcasting: adaptive sampling for sensor networks,

    R. Willett, A. Martin, and R. Nowak, “Backcasting: adaptive sampling for sensor networks,”Proceedings of the 3rd international symposium on Information processing in sensor networks (IPSN), pp. 124–133, 2004

    work page 2004

  19. [19]

    Machine Learning in Wireless Sensor Networks: Algorithms, Strategies, and Applications,

    M. A. Alsheikh, S. Lin, D. Niyato, and H. P. Tan, “Machine Learning in Wireless Sensor Networks: Algorithms, Strategies, and Applications,” IEEE Communications Surveys & Tutorials, vol. 16, no. 4, pp. 1996– 2018, 2014

    work page 1996

  20. [20]

    Privacy- preserving context-based electric vehicle dispatching for energy schedul- ing in microgrids: An online learning approach,

    Y . Liu, P. Zhou, L. Yang, Y . Wu, Z. Xu, K. Liu, and X. Wang, “Privacy- preserving context-based electric vehicle dispatching for energy schedul- ing in microgrids: An online learning approach,”IEEE Transactions on Emerging Topics in Computational Intelligence, vol. 6, no. 3, pp. 462– 478, June 2022

    work page 2022

  21. [21]

    Learning-based adaptive sensor selection framework for multi-sensing WSN,

    S. Ghosh, S. De, S. Chatterjee, and M. Portmann, “Learning-based adaptive sensor selection framework for multi-sensing WSN,”IEEE Sensors Journal, vol. 21, no. 24, pp. 27979–27990, 2021

    work page 2021

  22. [22]

    UK Air Quality Reanalysis (AQREAN),

    R. M. Barratt, N. P. Young, G. W. Curley, and D. R. Derwent, “UK Air Quality Reanalysis (AQREAN),”CEDA Archive, 2020. [Online]. Available: https://catalogue.ceda.ac.uk/uuid/7d44ef2a9e9346e79863f53193db189e/ 6

    work page 2020