EMMA4Drive – Dynamic Human Model for More Safety and Comfort in Autonomous Vehicles

For many employees, it is an inviting vision of the future: driving to work in their own car and still using the journey time wisely – reading messages, checking emails or relaxing and enjoying the first coffee of the day.

New digital tools for research, development and validation of the technology are needed to understand the expectations of autonomous vehicle customers, strengthen their trust, and demonstrate safety. With the EMMA4Drive project, our researchers are further developing the dynamic human model EMMA and adapting it for use in fully or semi-autonomous vehicles.

Motion Sequences Instead of Quasi-Static Tests

Until now, human models have been used either in crash simulations to estimate the risk of injury or in ergonomics analyses. In crash analyses, detailed, computing time-intensive models are usually used for calculations in the millisecond range, which are not suitable for simulating dynamic driving maneuvers, as longer processes have to be simulated here. In contrast, human models for ergonomics analysis are based on the simplified kinematics of a multi-body model and have so far only allowed quasi-static investigations. Realistic postures and movements for new activities can only be modeled with great effort using these models. In addition, the influence of dynamic driving maneuvers on the body cannot be taken into account.

IPS IMMA digital human model in a relaxed position — © Fraunhofer ITWM
The digital human model IPS IMMA takes a short break in the vehicle. EMMA helps to answer the question of how to assess comfort under dynamic driving manoeuvres even in such relaxed postures.

Sit Better, Operate More Comfortably

»Our human model uses an optimization algorithm to automatically calculate new body postures and entire movement sequences over a longer time window with the associated muscle activities,« explains project manager Dr. Marius Obentheuer. »This means that the simulation model can also be used to investigate the effect of dynamic driving maneuvers on people and their (reaction) behaviour – e.g. when designing assistance systems or control algorithms for (semi-)autonomous driving.« EMMA4Drive thus enables a comparatively simple implementation of new movement patterns and an efficient virtual investigation of safety, comfort and ergonomics.

Further Information

Predecessor Project EMMA-CC

Ergonomic workplace design is important for product development and production planning in order to reduce injuries and illnesses. The MAVO project EMMA-CC has developed a digital human model that facilitates the creation of safe and healthy workplaces.

EMMA-CC stands for »Ergo-dynamic Moving Manikin with Cognitive Control.«

Sitting Correctly Through Machine Learning

Calculating the contact between a seated person and the vehicle seat is very complex – particularly due to the large number of possible deformations on the contact surface. Up to now, this contact has been calculated in complex Finite Element (FE) simulations, for example for crash scenarios. Our research partner at the University of Stuttgart developed an AI model in order to be able to model this physically realistic contact in EMMA. It is based on current machine learning methods and was trained with data from FE simulations. This allows contact forces to be predicted quickly and efficiently based on the seating position – a decisive advantage for motion prediction in EMMA.

Link With the THUMS^® Human Model

In order to be able to compare the results from EMMA directly with those from FE simulations, our researchers transferred the established THUMS^® human model into a multi-body model that can be used for EMMA. THUMS^® is also frequently used in crash analyses in industry, as well as by our research partner. This interface allows EMMA simulations to be continued or validated in FE environments. Movements generated in EMMA can also be reproduced in FE simulations with THUMS^®.

EMMA bone model on a car seat — © Uni Stuttgart
The EMMA bone model is based on established occupant simulation models – this means that the motion data obtained can be further processed in other models, for example in crash simulations.

EMMA Masters Dynamic Driving Maneuvers

To evaluate driving comfort, we simulate a variety of dynamic driving maneuvers in EMMA: lane changes, evasive maneuvers, starting, braking – and increasingly also new scenarios from autonomous driving, such as vehicle takeover in hazardous situations.

To this end, our researchers added functionalities to EMMA so that vehicle movements can be specified variably in the simulation: from simple linear acceleration profiles to measured reference data. The latter were integrated using the RODOS^® driving simulator – for example during an emergency evasive maneuver. Now, even vehicle movements can be optimized by EMMA itself.

Predicting Movements Without Prior Knowledge

If the person is not driving the vehicle themselves, for example as a passenger or in an autonomous vehicle, they do not know the future maneuvers. This brings with it new challenges. Until now, movements in EMMA have been calculated using optimization methods – always taking into account the complete driving maneuver. This method provides realistic predictions for experienced drivers who are prepared to react to movements. An approach that is desirable for simulating experienced drivers.

Now, however, movements that react to unknown accelerations should also be predicted. To achieve this, our researchers developed a mathematical methodology based on model predictive control. It decouples the planning of the movement from the actual future state of the vehicle and thus enables realistic reactions to unpredictable driving situations.

EMMA4Drive – Dynamic human model for greater safety and comfort in autonomous vehicles

EMMA4Drive focuses on the simulation of seated people in cars. However, other vehicle types are also relevant – such as buses, public transport and construction machinery. In an industrial project, we investigated the behavior of standing passengers in an autonomous shuttle. The video shows an example of a braking maneuver.

© Fraunhofer ITWM
A digital image of a female driver is used to model quasi-static sitting in the vehicle seat.

EMMA on RODOS^®

And before EMMA can take to the road, she must of course pass her driving test – virtually in our interactive driving simulator RODOS^® (RObot based Driving and Operation Simulator). However, a real person first takes a seat there to collect physical measurement data and validate the results of the simulation software against measured data. The interaction between the driver and the seat is examined, for example the pressure distribution. This data should help to better answer fundamental questions about autonomous or semi-autonomous driving: How quickly should the tilted backrest of a seat be raised again using the integrated electric motor system? Or return the rotated seat to its original position? How long does it take before a person can take the wheel again when the vehicle signals in semi-autonomous mode: »Danger from the right, please take over!«?

The driving simulator is a central component of the Technikum in the division »Mathematics for Vehicle Engineering« and allows the use of different production cabs and real car bodies mounted on a strong robot arm. Currently, the researchers are working on a combined biomechanical-mechatronic model of the coupled seat system, which can be used to parameterize and calibrate the simulation software developed in the EMMA4Drive project.

This means that in the future, certain tests that are primarily aimed at physically stressing the occupants can also be carried out purely virtually, in addition to individual RODOS^® simulator studies in a real driving cabin. When testing new concepts or comparative investigations of alternative variants, this saves time-consuming hardware modifications. However, for studies in which psychological aspects of the driving behavior are in the foreground, the simulation with RODOS^® in a realistic cabin environment remains indispensable, since it is crucial for achieving a perfect immersion of the human being into the driving situation.

Video of the »EMMA4Drive« Study in RODOS^®

We used our interactive driving simulator RODOS^® to investigate autonomous driving in a larger study. The focus here was on the seating comfort perceived by the test subjects and their perception of safety during an emergency evasive maneuver. The subjects were equipped with a »full body tracking system« and the seat surface was also fitted with pressure sensors. This measurement of body movements helps us to correlate the forces that occur during the evasive maneuver with the subjective perception of comfort.

Project Duration and Funding

The project »EMMA4Drive – Dynamic Human Model for Autonomous Driving« ran from April 2021 to December 2024 and was jointly funded by the German Research Foundation (DFG) and the Fraunhofer-Gesellschaft. With a funding volume of five million euros, the project supported the development of realistic human models for the design of (partially) autonomous vehicles. A key aim of the funding was to initiate the transfer of scientific innovations to industrial applications at an early stage – and to actively involve companies in the research.

Project Partners

EMMA4Drive / dynamic model — © Fraunhofer ITWM
The dynamic model simulates muscle regulation during targeted movements. In this way, comfort as well as the risk of injury to the occupants can be investigated.

EMMA Learns to Drive: Dynamic and Autonomous