Research project REMAP – Reliable Multimodal Adaptive Predictive Maintenance in a Changing World
How can we prevent heavy vehicles from breaking down? REMAP develops explainable AI that learns from complex workshop and sensor data, detects changes over time, and updates itself through feedback. The goal is reliable maintenance and sustainable transport.
Image: Zara Karazian.
Predictive maintenance helps prevent breakdowns by detecting problems before they occur. However, in the heavy-vehicle industry, turning such models into reliable real-world solutions remains challenging.
Workshop data are complex and fragmented, collected at irregular intervals, and often lack clear links between symptoms and repairs. In addition, vehicles operate under changing conditions, and failure data are limited.
This project develops robust and explainable AI methods that can learn from multimodal data such as sensor readings, service records, and text. By detecting data drift, updating models through feedback loops, and enriching data using large language models, the project enables predictive maintenance systems that adapt over time.
The goal is to create trustworthy, production-ready solutions that increase component utilization, support individualized maintenance planning, and contribute to a more sustainable and efficient transport system.
The REMAP project builds on the RAPIDS project and brings together Traton AB, Stockholm University, KTH and Linköping University. We develop robust machine learning methods for generating, monitoring, and updating predictive maintenance models from complex, multimodal, and drifting workshop data.
REMAP is funded by Vinnova, Transport and mobility solutions – FFI.