Transfer Learning: An AI assisted method to accelerate the passive safety validation

14 / 11 / 2024

Crash protection will be one of the 4 main focus points of the future EuroNCAP assessment plan due to raising requirements for the occupants’ safety as well as more sophisticated passive safety systems. To assess the performance of passive safety systems the vehicle size, the structural design and the restraint systems are crucial. Assuming a gaussian distribution of body shapes, mainly the 5^th, 50^th, 95^th percentiles crash test dummies are used to assess the interaction between the occupants and the vehicle [1]. Meanwhile, various research activities have shown differences in the frequency of a crash and the injuries in the event of a crash between body types, age and gender [2]. Virtual testing has always been a great counterpart to real world tests because it is faster and less expensive. However, simulated results still have to be validated through physical tests and come with high computational effort when using detailed virtual models. Additionally, the limited coverage of the diversity in human bodies is still an issue [1].

The AWARE2ALL project aims for a more equal safety assessment trying to include the vast variety of different human body shapes and types. Utilizing Machine Learning algorithms can help accelerating simulations, improving the quality of the results and including body diversity, as the simulation of different body shapes is a time-consuming process. That’s why the project partners in AWARE2ALL are investigating innovative AI based approaches for acceleration of the assessment for different body shapes (see Figure 1).

Figure 1: Overview of the planned transfer learning model

THI deploys its database with more than 30 000 different crash configurations to train the machine learning methods (DOI: 10.5281/zenodo.13998257). Next step, which is currently being investigated deals with the question how to profit from the already trained system for a more complex representation of a human body. In order to complete this task, the partners THI and ESI collaborated within the AWARE2ALL project. The THI crash model, which was developed in the simulation package LS-Dyna was converted to the ESI simulation package VPS. Next, the antropomorphic testing device (crash test dummy) Hybrid III was replaced by the human body model (VIRTHUMAN), which is available only in VPS. The partners generated a database, which is being currently deployed to find an optimal way for the injury level prediction for different crash test configurations (DOI: 10.5281/zenodo.14000200).

Subscribe to our newsletter to not miss any update on the project results and follow us on social media.

[1] F. Plaschkies, O. Vaculín and A. Schumacher. Assessment of the Influence of Human Body Diversity on Passive Safety Systems: A State-of-the-art Overview. Proceedings of the FISITA 2021 World Congress, Prague, 13–17 September 2021. DOI: https://doi.org/10.46720/F2021-PIF-071.

[2] Euro NCAP. Euro NCAP Vision 2030. A Safer Future for Mobility. Report. 2022.

News

DiverSim: an Inclusive Simulator for Diverse Synthetic Pedestrian Data Generation

25 / 10 / 24

One of the critical challenges in autonomous driving and AI-based systems is the lack of diverse and representative datasets. Existing datasets often fail to reflect the broad range of pedestrians found in real-world scenarios—especially underrepresented groups such as people with disabilities or individuals from various ethnic minorities. This gap in representation can lead to the […]

News

Improving Pedestrian-Vehicle Communication with Acoustic and Visual Interfaces (eHMI)

27 / 09 / 24

The primary goal of AWARE2ALL is to tackle the new safety issues that arise with the introduction of highly autonomous vehicles in mixed traffic environments. This will be achieved by developing inclusive and innovative safety systems (both passive and active) and Human-Machine Interaction (HMI) systems, both external and internal. These systems will consider the diverse […]

Resources

Lateral Evasive Maneuver with Shared Control Algorithm: A Simulator Study.

3 / 07 / 24

In Sensors Journal, was presented this paper introducing Lateral Evasive Maneuver with shared control algorithm to the world. Don't miss the chance to read more about AWARE2ALL and see its evolution.