Smart city

Overview

A smart city uses data from sensors, public systems and residents to run urban services better. Traffic signals adjust to actual flows. Bus arrival times are predicted from live positions. Street lights dim when streets are empty. Waste collection follows how full the bins are rather than a fixed route. Planners decide where to build cycle lanes or schools from evidence about how people actually move and live.

The technology is only part of the picture. A city is a public institution with residents who did not choose to be measured, so smart city work involves governance, privacy, procurement, accessibility and public trust as much as sensors and models. The most successful programmes start from specific public problems, such as congestion, air quality or slow permit processes, and use data to solve them in ways residents can see and understand.

An everyday picture is a large apartment building with a responsive caretaker. A good caretaker notices the lift is slow before residents complain, knows which pipes leak in winter and schedules repairs before damage spreads. A city with good data and good processes can behave the same way at the scale of streets, networks and neighbourhoods.

What it is

Behind these uses sit common building blocks: connected devices and networks, a city data platform that collects and shares data, open data portals, analytics and AI, and governance frameworks that set rules for what is collected and how it is used.

Where it came from

Cities have used data and automation for a long time. Computer-controlled traffic signal systems date back decades, and adaptive systems that adjust timings to measured traffic, such as SCOOT, developed in the United Kingdom and described in the early 1980s, are still widely used. Geographic information systems became standard tools in planning departments from the 1980s and 1990s.

The phrase "smart city" became widespread in the late 2000s and 2010s, when mobile networks, cheap sensors, smartphones and cloud computing made it possible to collect city-wide data continuously. Technology companies, governments and universities launched large programmes, and some ambitious new-build projects were proposed. Experience showed that technology-led projects often struggled with public acceptance, privacy concerns and long-term maintenance.

Standards and frameworks followed. ISO 37120, first published in 2014, defined indicators for city services and quality of life, so cities could measure and compare progress. Privacy law such as the GDPR shaped how personal data may be collected in public spaces, and AI rules such as the EU AI Act restrict certain uses, such as some forms of remote biometric identification. Today the emphasis has shifted towards resident-centred, problem-led programmes, open data and digital twins of urban systems.

How it works

Step 1: start from a public problem

Pick a problem residents feel, with a measurable outcome: average bus delay, water lost to leaks, response time to service requests, or days with poor air quality.

Step 2: gather and connect data

Combine existing data, which cities often already hold in many departments, with new sensors where needed. Low-power wide-area networks and mobile networks connect devices. Data flow into a city data platform, often through streaming systems such as Apache Kafka, with common identifiers for streets, assets and areas.

Step 3: protect privacy from the start

Collect only what the purpose needs. Aggregate or anonymise data early, for example counting people rather than tracking them, and remove personal details from records with tools such as Presidio. Run data protection impact assessments and publish clear notices.

Step 4: analyse and model

Typical methods include time-series forecasting for demand and travel times, anomaly detection for leaks and faults, optimisation for routes and signal timings, computer vision for counting vehicles and pedestrians, and spatial analysis that combines many map layers. Some cities use federated learning or on-device processing to keep raw data local.

Step 5: act and inform

Feed results into operations, such as signal control or crew dispatch, into planning decisions, and into public information such as live transit apps and open dashboards. Many uses are real-time systems with strict latency needs.

Step 6: evaluate and govern

Measure the outcome against the baseline, publish results, and review algorithms for bias and unintended effects. Keep a register of the algorithms the city uses and who is accountable for each.

Key concepts and terms

A worked example

This scenario is illustrative. A mid-sized city loses about a quarter of its treated water through leaks in old pipes. Crews find leaks mainly when water appears at the surface.

The water department installs pressure and flow sensors at 60 points that divide the network into district metered areas. Night-time flow, when legitimate use is lowest, is a good indicator of leakage. A forecasting model learns each district's normal night flow given season and temperature, and an anomaly score rises when actual flow exceeds the forecast for several nights.

In the first year the system flags 34 districts. Crews with acoustic equipment search those districts first and find hidden leaks in 27. The model's alerts let crews spend most of their time where leaks are likely instead of surveying the whole network. Water loss falls by several percentage points over two years, and the city publishes the district-level results on its open data portal. No personal data are needed at any stage, which simplifies governance.

Strengths and limitations

Strengths:

• Improves service quality and efficiency in transport, water, energy and waste.
• Gives planners evidence about how the city is actually used.
• Enables faster detection of faults and incidents.
• Open data can support research, start-ups and civic projects.
• Helps measure progress on climate and quality-of-life goals.

Limitations:

• Surveillance and privacy risks can erode public trust.
• Vendor lock-in and incompatible systems raise long-term costs.
• Sensors and platforms need maintenance budgets beyond pilot funding.
• Data and models can reflect and reinforce existing inequalities.
• Benefits can be hard to measure and attribute.

How it compares with alternatives

Practical guidance and best practices

1. Start from a specific public problem with a measurable outcome.
2. Use data the city already holds before buying new sensors.
3. Design for privacy: collect the minimum, aggregate early and explain clearly.
4. Require open standards and data portability in procurement.
5. Budget for maintenance and replacement, not just installation.
6. Involve residents and front-line staff in design and evaluation.
7. Publish algorithms, data and outcomes where possible.
8. Check models for uneven performance across neighbourhoods and groups.

Common mistakes

• Launching pilots without a plan to scale or maintain them.
• Collecting identifiable data, such as faces or device identifiers, when counts would do.
• Locking the city into one vendor's closed platform.
• Measuring success by the number of sensors installed rather than outcomes.
• Ignoring areas with poor connectivity or residents without smartphones.
• Treating data from complaints and apps as representative of all residents.

Frequently asked questions

What makes a city "smart"?

Not the number of sensors. A smart city uses data and technology to improve outcomes residents care about, such as shorter journeys, cleaner air, reliable water and faster services, and does so transparently and fairly. Many effective projects use modest technology well.

Is a smart city a surveillance city?

It does not have to be. Many valuable uses need no personal data at all, such as pipe pressure or bin fill levels. Where people are involved, counting rather than identifying, aggregating early, strict retention limits and public oversight can keep benefits while protecting privacy. Laws such as the GDPR set binding limits.

Where does AI help most?

Forecasting demand and travel times, detecting faults and anomalies, optimising routes and signal timings, counting with computer vision, and helping staff and residents find information. High-stakes uses, such as policing or benefits decisions, need much stronger safeguards and are restricted in some jurisdictions.

How do cities pay for smart city projects?

Through a mix of capital budgets, national or regional grants, utility revenue, savings from efficiency, and partnerships. A common lesson is to include long-term maintenance costs from the start, because many pilots fail after grant funding ends.

What role does open data play?

Open data lets researchers, companies and residents build tools, check city performance and suggest improvements. It also improves internal data quality, because publishing data exposes errors. Personal or sensitive data must be removed or aggregated before release.

Glossary

• Smart city: a city that uses data and technology to improve services and outcomes.
• IoT: the internet of things, networked sensors and devices.
• GIS: geographic information systems for spatial data.
• Open data portal: a website where a city publishes datasets.
• District metered area: a section of a water network with measured inflow and outflow.
• Interoperability: systems working together through shared standards.
• Data minimisation: collecting only data needed for a stated purpose.
• Algorithm register: a public record of algorithms used by a government.

Summary

A smart city uses sensors, data platforms and AI to improve transport, water, energy, environment and public services, and to plan with evidence. The idea grew from decades of traffic control and GIS into city-wide data programmes in the 2010s, with standards such as ISO 37120 and privacy law shaping practice. The best programmes start from resident problems, use data the city already has, protect privacy by design, avoid lock-in and fund long-term maintenance. Technology helps only when governance and public trust keep pace.

Related

It belongs to Industries and applications in my stack, under Machine learning. Others in the same group: