ATTENTION - Artificial intelligence for real-time injury prediction

The aim of the ATTENTION research project is to develop a method for predicting injuries to vulnerable road users (VRU) such as pedestrians or cyclists in real time, based on machine learning methods.

With the help of data-driven procedures and digital human models, video data from vehicles and virtual tests are used to determine a situation-specific risk of injury. In the future, injury prediction could help to ensure safe and efficient transport thanks to risk minimization strategies for automated vehicles.

Key achievements as a result of the project include putting in place a position and movement database for pedestrians and cyclists. Data from Bosch accident research are analyzed and used as a basis for biomechanical and AI-based movement prediction. Finite-element crash simulations then provide potential injury patterns, which are compared with real accident data and stored in an additional collision and injury database. The results of the crash simulations serve as training data and lay the foundation for an AI prediction of situation-specific injury values, from which an injury risk index is derived. Based on the injury risk index, further measures will be defined, such as an adaptation of the driving behavior. The realization of the project goals will be done using a virtual demonstrator.

DYNAmore has been very successful in the area of human body modelling (HBM) for many years and will mainly cover the training data generation and the development of a machine-learning-based HBM model positioning approach. This will eventually be the basis for various human body model crash simulations. The model positioning approach will be realized using model-order-reduction techniques and enables the application of time-optimized positioning methods for human body models, which are based on actual simulations. Hence physical dependencies are accounted for during the model positioning, like contact conditions or material stiffness distributions, while taking only a fraction of the time a full body simulation would require.

The project is funded by the Federal Ministry for Economic Affairs and Energy (BMWi) and is managed by the German TÜV Rheinland. Other project partners are the Robert Bosch GmbH and QualityMinds GmbH. On the research side, the Fraunhofer Institute for High-Speed Dynamics (EMI) and the Institute for the Modelling and Simulation of Biomechanical Systems of the University of Stuttgart will participate in the project and Mercedes-Benz AG will contribute as an associated partner.

Attention_2.png

Project duration: 07/2021 – 06/2024

Funding program: Federal Ministry for Economic Affairs and Energy (BMWi);

Research program: “New vehicle and system technology”

Consortium partners:

  • DYNAmore (Gesellschaft für FEM Ingenieurdienstleistungen mbH)
  • Fraunhofer-Institute for High-Speed Dynamics, Ernst-Mach-Institut, EMI
  • QualityMinds GmbH
  • Robert Bosch GmbH
  • University of Stuttgart – Institute for the Modelling and Simulation of Biomechanical Systems (IMSB)

Associated partners: Mercedes-Benz AG