Foundation Models for Time-Series Automotive Large Scale Data

Purpose and goal:

In Autofreight 1 & 2, the test vehicles have been equipped with advanced logging equipment. This data will now be stored, managed and researched on more. Analyzing time series data helps us to detect critical events, since existing methods for event identification are inefficient pattern matching or machine learning that often are expensive or challenging to implement at scale.

In this project, Foundation Models will be tested for large data sets. Tire data, particle emissions and video data for traffic safety with A-double truck combinations will be analyzed. Expected results and effects:

Easier data sharing between research partners. Validate the potential of foundation models for analyze and generate plausible time-series data. Data that enables researchers to perform experiments, simulations and scenario analysis without relying solely on limited real-world data.

This may have the potential to pave the way for a whole new generation of digital twins. Research paper sharing insights on the performance in terms of accuracy and computational efficiency of the models in the context of traffic analysis and how it can be complemented with video data. Approach and implementation:

The planned work is structured into five work packages. The industrial project partners, Volvo and Goodyear are address industrially relevant innovation enablers to reduce the environmental impact of container transports. Their work is supported by fundamental and applied research in the area of foundation models for data analytics of time-series and unstructured data at a large scale driven by the academic project partners University of Gothenburg, Chalmers and VTI.