Center for AI

Safer fire department missions with advanced technology

Firefighters work under immense pressure when responding to a call. Extreme heat, smoke that obstructs vision, time pressure and danger are all factors under which they have to react. The stress created by such situations affects the ability to react, both physically and mentally, potentially having a serious impact on the cognitive capabilities. Those affected are often unaware of their limited ability to judge, which can have fatal consequences since every mistake can cost your own or someone else’s life.
Stressdetection Firefighter
Equipped with VR goggles and headphones, the testers undergo a mission.

To reduce this risk, IBM and fortiss joined forces to launch the “Stress Test for Firefighters” project at the Center for AI Research. The project partners have been working since late 2019 to find ways to measure and estimate the stress level of firefighters in real-time with the aim of helping them make decisions during a mission.

The Center for AI Research team comprises fortiss specialists from the fields of research and development, human-centered engineering and machine learning. For this project, the team also regularly relies on the crucial expertise of volunteer and professional firefighters. The project is thus following a design-thinking approach in which specialists from various areas jointly work on an issue, develop solutions, then continuously test and implement them.

The challenge of monitoring stress

For the underlying data, the project utilizes general stress indicators such as heart rate, brain activity, muscle tension, skin moisture or the level of the stress hormone coritsol in the blood. Another factor that has to be taken into account is the individual reaction to stress, which depends not only on the person’s current mental capacity, but also on the situation the person is in when the measurement is carried out.

Using this data as a foundation, fortiss and IBM are developing user-centric machine learning algorithms for stress monitoring based on data mining and cognitive characteristics. This allows to the team to develop new personalized stress recognition models by means of various firefighting scenarios and experience gained from such missions. The goal is to define a clearly understandable user stress state. Here the team relies on different machine learning approaches such as multi-class classification with a C5 classification tree model and “label-free” feature extractor, as well as an efficient personalization method thanks to the “human-in-the-loop” approach.

Important milestone

A design-thinking workshop was held in December 2019 as the first important project milestone. With the help of volunteer firefighters, among other things the IBM/fortiss team developed a storyboard that details all of the individual stations of a mission, which in turn clearly illustrates the individual stress situations, from the initial emergency call and drive to the site of the incident, to the post-mission debriefing. The team also worked together with the firefighters to develop various mission scenarios, which were then converted into a virtual reality demo.

  Marketing & press

Your contact

Marketing & press

presse@fortiss.org