Project ATTENTION – AI Makes Road Traffic Safer
QualityMinds supports the ATTENTION research project, which aims to reduce injury risks in road traffic based on AI.
Follow us
AI Research Project
ATTENTION
The 3-year project “ATTENTION”: Artificial Intelligence for Real-Time Injury Prediction” is funded by the German Federal Ministry for Economic Affairs and Energy (BMWi). It aims to enhance the safety of vulnerable road users (VRUs), such as pedestrians and cyclists, in automated traffic.
The project focuses on developing a real-time injury prediction method using machine learning algorithms and AI techniques. By analyzing vehicle-mounted video data and digital human models, the system determines injury risks. These predictions will support risk minimization strategies for automated vehicles, contributing to safer and more efficient traffic. The ultimate goal is to improve pedestrian and cyclist safety in automated transportation.
QualityMinds GmbH supported the 3-year research project ATTENTION in collaboration with various partners, including Robert Bosch GmbH, TÜV Rheinland, Fraunhofer Institute for High-Speed Dynamics (EMI), the University of Stuttgart, and Mercedes-Benz AG.
The Challenge:
Minimizing the injury risk of highly vulnerable road users in the event of an unavoidable collision.
This challenge is particularly relevant in autonomous driving, where last-minute decisions are required.
Urban traffic, characterized by limited space and diverse road users, still poses risks for unprotected road users such as pedestrians and cyclists, despite ongoing advancements in automation. To enhance safety in automated traffic, injuries among these vulnerable groups must be minimized.
AI-powered, data-driven methods are used to assess situation-specific injury risks by analyzing video data, virtual tests, and digital models.
The Solution:
QualityMinds combines Data Science and Machine Learning to enhance traffic safety.
QualityMinds GmbH supported the 3-year ATTENTION project in the selection and enhancement of test and training data, the design and implementation of Machine Learning (ML) models, and the development of software solutions to demonstrate the methods and findings.
The selection of appropriate data and features
The QualityMinds team was responsible for, among other tasks, selecting appropriate data and features. For data generation, the team analyzed numerous videos of relevant accident scenarios, such as dashcam footage from South Korea. Using Data Science, they extracted typical movement patterns and estimated human body postures from the video data.
The data included poses, skeletal data, and movement classes. Based on this, the team derived motion vectors, injury risks, and action recommendations for vehicle control.
Key Facts at a Glance:
Selection, Enhancement, and Generation of Test and Training Data
Analysis of Relevant Features for Predicting Follow-up Kinetics using Data Science
Development of a Visualization for Injury Risks Based on Open-Source Real-Time Simulation for Road Traffic (e.g., CARLA)
Calculation of Correction Proposals to Reduce Injury Probabilities
Improvement of Prediction Quality of Simulation Results through ML Models
Results: Creation of motion vectors, calculation of injury risks, and development of action recommendations for vehicle control
Project Results:ATTENTION
Development of a Position and Movement Database for Pedestrians and Cyclists (Unprotected Road Users)
Use of Crash Simulations to Identify Potential Injury Patterns
Comparison of Simulated Injury Patterns with Real Accident Data and Storage in a Database
Creation of an Injury Risk Index Using AI Techniques
Derivation of Measures Based on the Injury Risk Index
Implementation of the Findings in a Virtual Demonstrator
“The collaboration with partners from various sectors such as industry and research provided us with valuable insights. What was also great is that we were involved in every step of the research process, from data analysis and AI training to the final software with the virtual simulator.
Thanks to the project, we now have a powerful on-premise AI cluster in our office in Nuremberg, which enables efficient AI training and an agile organization of our work, allowing us to easily try out new ideas.”
Niels Heller, qualityminds
Project Manager & Machine Learning Engineer