Enhancing eCall Systems with LLM-Powered First Aid Guidance and Follow-up Information

Anamaria Dumitrescu, Cristian-Alexandru Tanase, Matias Vaduva, Bogdan Mocanu

Abstract: Since 2018, eCall has been mandatory in all new vehicles within the European Union, providing emergency services with a minimum data set in the event of a crash. While this system improves response times, its current design is limited to transmitting essential but sparse information. The objective of this work is to enhance eCall by delivering richer, incident-specific details that improve both situational awareness for first responders and immediate support for occupants. To achieve this, we integrate large language model (i.e., Gemma 2) into the vehicle as a first aid assistant. Using speech-to-text and text-to-speech interfaces, the assistant communicates directly with occupants to provide real-time guidance and simultaneously compiles a secondary follow-up message containing injury details and accident context. This information is transmitted over legacy 2G channels if a voice link with emergency services cannot be established. The innovation lies in repurposing the vehicle’s existing onboard computational resources to run advanced language models, extending eCall from a passive alerting mechanism into an active, occupant-supporting system. This next generation eCall framework not only increases the amount and quality of data available to emergency responders but also bridges the critical gap between crash occurrence and professional medical assistance, with the potential to significantly improve survival and recovery outcomes.

External IDs:dblp:conf/wpmc/DumitrescuTVM25