What is it about?
This paper is about making city drive-by environmental monitoring more reliable and fair when sensors have limited energy. We use solar-powered sensors mounted on taxis in Stockholm to measure street-level conditions (air quality, temperature, noise, thermal imagery), but taxi routes naturally oversample busy corridors while leaving other areas under-measured, and battery limits cause dropouts. The core idea is a simple smart on/off rule: if a location was sampled recently (on the order of minutes), the sensor can temporarily stay off there, saving energy and avoiding redundant measurements, and then focus effort when the vehicle reaches places that have not been sampled in a while, guided by health-relevant averaging times. We also show why you can still trust the big picture even with selective sampling: over time, the randomness of urban mobility averages out, so aggregated data converges to a stable city-wide pattern. In a five-month Stockholm deployment (five devices; 59,476 measurements), the adaptive strategy improved energy efficiency by ~10.7% and made coverage much more even (about 72% lower spatial variability), effectively shifting measurements from oversampled streets toward undersampled areas.
Featured Image
Photo by Leo_Visions on Unsplash
Why is it important?
Urban environmental risks are everywhere but uneven. Air pollution, noise, and heat can vary block by block, and cities need high-resolution monitoring to find hotspots and act quickly. Traditional fixed stations often can not capture that street-level detail, while mobile sensing can, but only if it is not biased toward the same few busy corridors. What is unique (and timely) here is that the method is designed for the messy reality of real fleets: routes are not controlled by researchers, there’s no central coordination, and devices face tight energy and bandwidth limits. These conditions make many classic sensor-network approaches impractical. Instead of trying to "move sensors to the right places", we adapt when to measure, using public-health guidance, specifically WHO-linked averaging windows, to cut redundant measurements and naturally shift effort toward under-measured areas. The other differentiator is trustworthiness: more than merely proposing a heuristic, the work also explains why long-term maps remain reliable even with selective sampling, grounding the approach in ergodicity and providing a formal convergence argument for time-averaged data. In practical terms, this can help cities build more equitable, lower-cost monitoring with small fleets and energy-limited devices, thus supporting better targeting of interventions and stronger environmental accountability as urban health pressures grow.
Perspectives
From a public-interest lens, this publication can be read as an argument that street-level environmental data should be treated like civic infrastructure. Something cities proactively build, maintain, and democratize, rather than a luxury produced by sparse and costly stations. Because drive-by sensing with taxis naturally creates data-rich main arteries and data-poor side streets, the absence of measurements can quietly translate into political invisibility. It becomes easier to ignore localized hotspots, delay interventions, or debate whether a problem really exists. The contribution here is a practical way to make monitoring more equitable and sustainable under real constraints. In that framing, the work is also about shifting power toward accountability, by reducing data deserts that often track where attention and investment are already missing.
mayar ariss
Massachusetts Institute of Technology
Read the Original
This page is a summary of: Ergodicity-Informed Adaptive Sensing for Energy-Constrained Urban IoT Networks, IEEE Internet of Things Journal, January 2025, Institute of Electrical & Electronics Engineers (IEEE),
DOI: 10.1109/jiot.2025.3621047.
You can read the full text:
Contributors
The following have contributed to this page
