Alstom is conducting extensive tests in collaboration with the Swedish startup Flox Intelligence on an AI system that detects and removes wild animals from the vicinity of railway lines.
The technology is based on artificial intelligence. Initial results show that the system has high potential to reduce the risk of wildlife collisions, one of the most common causes of rail disruptions in Sweden.
“Our field tests show that artificial intelligence can identify animals with a high degree of accuracy. By combining our experience in wildlife protection solutions with Alstom’s expertise in rail innovation, we are developing technology that protects both animals and the operational reliability of the network,” said Sara Nozkova, CEO of Flox Intelligence.
“The project represents an important step for the safety and sustainability of Swedish rail transport. According to her, the teams involved were surprised by the large number of animals observed even over short distances, particularly on sections where accidents frequently occurred,” said Maria Signal Martebo, CEO of Alstom Sweden.
The AI system detects animals in real time
The technology developed by the two companies is based on AI-assisted video cameras that identify animals in real time and activate tailored audio signals to scare them away from the tracks. In the first phase of testing, the AI system identified several species, including moose, deer, foxes, and wild boars. As of April 2026, the project has entered a new implementation phase that includes both video detection and acoustic warning systems.
The tests are being conducted in collaboration with the regional rail transport authority Tåg i Bergslagen and the operator VR, on several lines in Sweden, including the Dalabanan and Bergslagsbanan. The project also receives funding from the Swedish innovation agency Vinnova and is considered an important step toward the safety and sustainability goals promoted by the European Union.
The AI system detects not only large animals but also smaller species and birds that previously appeared almost never in railway statistics. Each identification is analyzed and classified, allowing the algorithms to continuously improve their performance. During testing, the system proved particularly accurate in detecting domestic animals and birds such as crows or pigeons, while identifying moose and deer required an additional period of AI training.
Every year, approximately 5,000 train-animal collisions are reported in Sweden. In addition to the impact on wildlife, these incidents cause delays, repair costs, and have a psychological impact on train engineers. By reducing the number of accidents, the project aims to improve punctuality, reduce environmental impact, and improve working conditions in the railway sector.
Share on:
