The Dutch city of Heerlen is receiving €5 million in EU funding for an algorithmic surveillance system based on a “digital twin”. Frontex is also interested in the technology.

The Dutch municipality of Heerlen is using €5 million in funding from the European Union to build an AI-supported system designed to predict crime and disturbances of public order. The project, launched last year, is called “Pulse-Twin”. An initial version is expected to be completed later this year and will then be tested until spring 2028.

Predictive software for police authorities has been tested worldwide for around two decades. Previous systems primarily focused on forecasting specific offences such as burglaries or theft. Artificial intelligence is intended to take these approaches to a new level.

The technical basis is a so-called digital twin – a continuously updated digital replica of the real city, incorporating data from social, statistical and police sources. An algorithm processes this information and calculates probabilities for disturbances of public order or criminal offences. Based on these forecasts, authorities are expected to be able to act “preventively”. In addition to crime prediction, according to those responsible for the project the system can also be used for traffic management and energy management in buildings.

World’s first project in an urban environment

The EU funds for “Pulse-Twin” come from the European COVID recovery fund. According to its operators, the project aims to address “urban challenges with data-driven solutions”. Heerlen is considered the first city in the world to introduce such a system.

The Dutch section of Amnesty International criticises the project: “Systems that predict nuisance or crime and are used for law enforcement by the police carry a high risk of ethnic profiling. Amnesty International therefore hopes that the municipality of Heerlen will reconsider the use of Pulse-Twin.”

This warning is based on a structural problem of algorithmic systems that rely on historical police data: if certain population groups or neighbourhoods were disproportionately targeted in the past, the model reproduces these patterns.

This is also the criticism of systems such as Gotham by Palantir, which have already been introduced or are planned in four German federal states. According to authorities and the US manufacturer, however, Gotham cannot access data on the internet and does not operate according to the principle of the digital twin.

Frontex invites manufacturers to “industry day”

The concept of the digital twin has long also been used for decision-making and command in the military sphere. As a virtual representation of, for example, tanks, aircraft or troop formations, AI links sensor data in order to simulate operations before a weapon is deployed or to predict problems in supply chains. It also optimises military training and maintenance requirements.

The EU border agency Frontex is also working on the concept. As part of its Copernicus border surveillance service, the agency is developing digital twin models of the EU’s external borders. These are intended to enable officials to simulate specific scenarios – such as irregular border crossings under different weather conditions.

Frontex aims to deploy resources more effectively and make surveillance technology more efficient. The system is intended to be geographically scalable and capable of modelling different types of borders, from river courses and coastal sections to forest areas and mountainous regions.

Last June, Frontex held a two-day competition with a total budget of €180,000, in which three companies – Kondorcad Landsurv from Romania, GMV from Spain, and the Institute of Geodesy and Cartography from Poland – presented “digital twin solutions” for a terrain area near Dubova in Romania. For the potential procurement of digital twins for border forces in EU member states, Frontex invited European manufacturers to an “industry day” in February.

Image: Digital twin of a border region (Frontex)