Signal Atlas

Research library

Digital twins & simulation

Live virtual copies of cities, factories and networks, used to test a change before making it for real. What 6G adds: A twin is only useful if it stays in step with the real thing. 6G plans built-in sensing, edge AI and low delay to keep it in step, and operators use network twins to plan 6G itself.

Of 34 findings in this theme, 21 are real today, 9 designed and 4 forecast.

Real today 21

  • A real digital twin stays linked to its target. A digital twin is a computer model kept in step with a real thing (a machine, network, building or city) through a continuing data link, so it can be used to test 'what if' before acting. The owner's notes and the national science academies agree the live, two-way link is what separates a twin from an ordinary model.
  • BMW plans whole factories in a virtual twin. BMW uses NVIDIA Omniverse to plan factories virtually, testing layouts, logistics and robot placement before construction; its Debrecen plant is the showcase example.
  • Chattanooga's traffic twin estimated 16% energy savings. US national labs built a twin of Chattanooga traffic from live cameras and radar to test signal changes before applying them. Retiming lights on one congested corridor was estimated to save up to 16% energy; the city-wide goal was 20%.
  • Cities and nations are building live digital twins. Singapore pioneered a full digital double of the city, the UK set up a National Digital Twin effort in 2018, and Rotterdam and Amaravati (India) are building their own. These twins need pervasive sensors, but IoT lacks a common protocol.
  • Digital twins have a standard vocabulary; 6G adds sensing. ISO/IEC 30173 (2023) defines digital twin terms and concepts. A 2025 Next G Alliance paper argues 6G sensing and AI will make network, factory and city digital twins more accurate.
  • Dubai launched an emirate-wide twin in July 2026. Dubai Municipality launched a Digital Twin Platform with 3D models of more than 195,000 buildings and 280,000 infrastructure assets, for planning, asset management and simulations.
  • EU is building a digital twin of Earth. The EU's Destination Earth programme ran its first two twins (climate adaptation and weather extremes) from 2024 and aims for a 'full' digital replica of the Earth by 2030 to support climate policy. The owner's notes cite it as the big bet on policy-simulation twins.
  • Factories have a twin standard since 2021. ISO 23247 (2021) is a framework for factory digital twins, covering synchronisation, data collection, simulation and security. NIST says small manufacturers still struggle and is starting a part on verifying twin accuracy.
  • Free simulators let cities test traffic before changing roads. SUMO, an open-source traffic simulator from the German Aerospace Center since 2001, produced traffic forecasts for Cologne during the 2005 papal visit and 2006 World Cup. CARLA (2017) is a free urban driving simulator for self-driving research.
  • Helsinki publishes its city model as free open data. Helsinki's '3D' twin combines a reality mesh accurate to about 20 cm and a building data model, released openly; it powers an Energy and Climate Atlas estimating heating demand and solar potential. A 2018-2019 Kalasatama project built twin models as a test platform.
  • Japan opened 3D city models for 300+ places. Japan's national PLATEAU project publishes standard 3D city models as open data; the ministry's list counted 306 locations in May 2026, up from 56 cities in 2021. These models are the base layer other twins can build on.
  • Many city 'twins' are 3D models without live data. A lot of what is marketed as a city digital twin is a detailed 3D map updated once a year or 'when required', with no live feed and no link back to decisions. Zurich's own planners describe their twin as 3D spatial data updated at varying frequency, and the owner's research separates static models, one-way 'shadows' and true closed-loop twins.
  • Ottawa built a planning twin with heat maps. Since about 2022 Ottawa has built a photorealistic city twin from imagery, laser scans and VR, used to show how new towers shade homes and parks and to map the hottest areas for heat-wave planning with public health.
  • Robots and self-driving cars learn in simulation first. Robots and self-driving cars are now routinely trained and tested in simulation. A 2017 study trained a robot's vision only on simulated images and it worked in the real world to about 1.5 cm; NVIDIA's open-source Isaac Sim and Waymo's simulator do this at industrial scale.
  • Shanghai district claims China's first urban twin. Shanghai's Lin-gang area describes a twin mapping the district one-to-one, used to simulate a major venue before construction. It aimed to finish the overall structure of its 'twin city' by 2025, part of China's national push for city twins.
  • Small German town tested a twin with residents. Supercomputing researchers in Stuttgart built a twin of Herrenberg combining a 3D laser scan, traffic data, airflow simulation and an app where residents marked places that felt unsafe or ugly, shown in virtual reality at public events.
  • Toronto shares a yearly 3D model, not a twin. Toronto's City Planning publishes a city-wide 3D 'massing' model of buildings as open data, refreshed about once a year (latest December 2025). It is useful for planning and apps but is a snapshot, not a live twin.
  • Twins of the mobile network itself are being standardised. A 'network digital twin' is a live virtual copy of a telecom network used to test changes, plan capacity and train AI safely before touching the real network. ITU-T published requirements and an architecture for it in 2022 (Y.3090), and the 6G framework says such twins can help verify, simulate, deploy and manage 6G networks.
  • Vendors sell simulators of whole 5G/6G city networks. NVIDIA offers a platform for building physically accurate virtual 5G and 6G networks, from one tower to a whole city, as part of its 6G developer program. Nokia's 6G white paper describes using such geo-spatial twins to improve radio network performance.
  • Virtual Singapore: a national 3D twin built 2014-2018. Singapore launched Virtual Singapore in December 2014, an S$73 million research programme to build a detailed, 'semantic' 3D model of the whole country able to take live sensor data. Planned uses included solar potential, flood modelling and testing ideas before building.
  • Zurich uses its twin for planning and climate. Zurich's administration describes a city twin built from open 3D data, used to show new buildings in context, run architecture competitions, support public participation and simulate urban climate questions.

Designed, not deployed 9

  • 3GPP finished a study on network twins in 2026. The body that writes mobile standards completed a Release 19 study on managing network digital twins (published by ETSI in January 2026). It lists uses such as predicting failures, checking energy-saving settings before rollout, and generating training data for AI, and recommends writing actual specifications next.
  • Connected twins need shared definitions to interoperate. Patient twins may exchange data with device, organ, hospital or emergency-service models. That needs shared units, time references, identities, versions and provenance, or one wrong state spreads through every connected model.
  • Ericsson shows big gains from twins, in simulation. Ericsson's 2022 review of network twins lists uses from site design to testing AI before rollout. In a simulated factory, twin-guided radio settings gave far higher data efficiency and slightly higher reliability than conventional methods.
  • ITU's 6G framework lists digital twins as a use. The UN telecom agency's official 6G framework (IMT-2030, 2023) expects 6G to help copy the physical world into precise real-time digital twins for industries such as health care, farming and construction. It also lists sensing data for digital twins as one job of 6G's network-sensing scenario.
  • Internet researchers say network twins are still early. An Internet research group draft (2026 revision) describes a network digital twin as a real-time, interactive copy of a data network used for 'what if' testing, training and planning. It says research is still in its infancy and lists open problems: handling huge data, matching reality closely, real-time sync and security.
  • Many twins can be stacked into city views. Standards material notes that digital twins of individual assets can be aggregated into composite views of ships, bridges, factories, supply chains or whole cities. In health, the same step could support hospital operations or public-health monitoring.
  • Open RAN group calls RAN twins key for 6G. In 2024 the O-RAN Alliance's research group, with contributors including Bell Canada, NVIDIA, Ericsson and Nokia, concluded that twins of the radio network could help 6G with AI training, energy saving, planning and site-specific tuning. A US university testbed funded through that group (2025) now automates data collection and testing for network AI apps.
  • Rotterdam builds a city-owned twin with ethics board. Rotterdam chose to lead and co-fund its own Open Urban Platform rather than leave city data infrastructure to big tech, with a planned governance board to vet commercial uses of the data. A digital twin is meant to grow on top of it.
  • UK set public-good principles for national twins. The UK's Gemini Principles (2018) set nine values for a national twin, from public good and security to openness and federation. The centre that wrote them closed in 2022.

Forecast 4

  • 6G sensing could keep twins of the world updated. Nokia Bell Labs and the North American Next G Alliance argue that 6G radios will also act like radar, sensing objects and movement, and that this data plus in-network AI could keep network, factory and city twins accurate in near real time.
  • A whole transport system twin is foreseen before 2050. The IEEE paper envisions a '360 degree' digital twin of an entire country's transport, linking every vehicle, operator and city system to optimise movement of people and goods in real time. It says the tools exist to do this before 2050 if stakeholders cooperate.
  • Foxconn expects 30% energy cut from factory twin. Foxconn built a twin of a new Mexican server factory to lay out lines and train robot arms and mobile robots virtually, and expects electricity use to fall over 30% a year.
  • Next step: many twins linked into ecosystems. The owner's notes predict 'systems of twins' that talk to each other, such as a building twin linked to grid and weather twins. Public bodies point the same way: the UK's plan describes a national twin as connected twins sharing data securely, and 3GPP's study includes twins nested inside twins.

Other themes