Computationally Efficient Human Body Modelling for Real Time Motion Comfort Assessment

Raj Desai, Marko M. Cvetkovic, Junda Wu, Georgios Papaioannou, Riender Happee

Introduction

As the automotive industry transitions toward high-level automated driving (AD), the interior cabin experience is undergoing a fundamental shift. Relieved of primary driving tasks, passengers are increasingly likely to engage in non-driving activities such as working, reading, or resting. These activities involve non-standard postures that increase susceptibility to whole-body vibration (WBV), which is a primary contributor to physical fatigue and chronic musculoskeletal issues like back pain.

To optimize vehicle suspension and seat design for these new use cases, engineers require robust tools to simulate human dynamic responses. However, existing modeling approaches present a significant trade-off. Simple 2D models are computationally light but fail to capture the lateral and rotational dynamics essential for modern comfort assessment. Conversely, high-fidelity models—such as the industry-standard MADYMO Active Human Model (AHM)—provide immense detail but are computationally prohibitive, often requiring hours of processing time for just a few seconds of simulation. This paper presents the "Efficient Human Model" (EHM), a 3D multibody approach designed to bridge this gap by providing real-time performance without sacrificing essential kinematic accuracy.

Methodology

The Efficient Human Model (EHM) Architecture

The EHM was developed using a multibody dynamics framework, striking a balance between anatomical detail and computational speed. The model consists of 12 rigid segments: the pelvis, three torso sections (lower, middle, and upper), the neck, the head, and the limbs. These segments are connected by a total of 31 degrees of freedom (DoF).

Key innovations in the EHM’s joint configuration include:

  • Strategic Joint Selection: The model utilizes a combination of spherical, universal, and revolute joints to mimic human range of motion.
  • Spinal Compression: A critical feature is the inclusion of a spherical-translational joint at the T12 vertebra. This allows the model to account for vertical spinal compression, a factor frequently omitted in simplified models but vital for capturing resonance under vertical vibration.
  • Active Postural Control: To prevent the "slumping" or postural drift common in passive models, the researchers implemented PID (Proportional-Integral-Derivative) controllers. These controllers act as active stabilizers, mimicking muscle activity to maintain an upright posture during dynamic loading.

Simulation Environment

The model was implemented using a MATLAB-Simulink-MADYMO co-simulation pipeline. This setup allows for rapid parameter optimization and ensures that the model can be easily integrated into existing automotive engineering workflows.

Validation and Results

Benchmarking Performance

The EHM was validated against experimental data from human subjects exposed to random 3D vibrations (0.3 m/s² rms) in a vehicle mock-up. Its performance was also compared directly against the high-fidelity MADYMO AHM.

  • Computational Efficiency: The most striking result was the speed of execution. The EHM completed a 35-second simulation in approximately 30 seconds. In comparison, the high-fidelity AHM required 3 hours to process the same 35-second window. This represents a 360-fold increase in speed, enabling the EHM to run faster than real-time.
  • Kinematic Accuracy: Despite its simplified structure, the EHM’s predicted responses for the head and trunk closely aligned with experimental ground truth. In several frequency ranges, the EHM’s predictions were found to be as accurate as, or even more accurate than, the significantly more complex AHM.

The Role of Spinal Dynamics

Sensitivity analyses highlighted the importance of the T12 compression joint. When this joint was locked (simulating a rigid spine), the model produced an erroneous resonance peak at 8 Hz. This confirms that vertical flexibility is a non-negotiable requirement for any model intended to assess passenger comfort accurately.

Discussion and Future Work

The EHM’s ability to perform faster-than-real-time simulations opens new possibilities for active vehicle control. It could potentially be integrated into "smart" suspension systems that adjust damping in real-time based on the predicted motion of the occupants' heads and torsos.

However, some limitations remain. The current iteration of the EHM is tuned to a 50th-percentile male and does not yet account for the wide variance in human anthropometry or individual muscle tension levels. Future research will focus on incorporating individual biological feedback loops—such as vestibular and visual perception models—to further refine how the model perceives and reacts to motion. Additionally, expanding the model to cover a broader range of body types will be essential for universal comfort optimization in autonomous fleets.

Conclusion

The Efficient Human Model (EHM) demonstrates that high-fidelity simulation is not always a prerequisite for accurate comfort assessment. By focusing on key anatomical degrees of freedom and utilizing active stabilization, the EHM provides a computationally efficient tool that is ready for the real-time demands of next-generation vehicle design.

Summary updated

BibTeX
@inproceedings{desai2023computationally,
  author = {Raj Desai and Marko M. Cvetkovic and Junda Wu and Georgios Papaioannou and Riender Happee},
  title = {{Computationally Efficient Human Body Modelling for Real Time Motion Comfort Assessment}},
  booktitle = {DHM},
  year = {2023},
  doi = {10.1007/978-3-031-37848-5\_32},
}
Figure