Page 15 - IJAMD-1-3

International Journal of AI for
Materials and Design
Smart cockpit design with generative models

virtual environment. The psychological facilitates real-time references to the designer and engineer. The designer
monitoring of the user’s mental health status, providing a agent is responsible for generating the design solution and
non-invasive method for continuous psychological well- balancing esthetics and functionality. The engineer agent
being tracking. The biological model simulates specific ensures the manufacturability and structural integrity of
internal mechanisms at the molecular and cellular levels. the design solution. Finally, the UX analyst agent evaluates
Given the diversity of data and the accuracy of HDT the UX factors of the design and offers suggestions to the
models, HDT can contribute to personalized design, designer, thus facilitating the collaborative design process.
verification, human-machine interaction experiences, and The automatic design process begins with the user agent
the applicability of human-centric product design.
extracting the user’s inputting personalized demands, such
3.2. Main functions and specific implementation as the design style of the vehicle seat. The designer agent
then retrieves or generates a preliminary cockpit design
The main functions of the proposed personalized design based on these specific requirements, employing a retrieval
framework for smart vehicle cockpits are illustrated in model, generative adversarial networks (GANs), or other
Figure 2. These functions are categorized into four main generative models to ensure design variety and innovation.
areas: (i) Personalized product design; (ii) personalized The engineer agent then evaluates the manufacturability
interactive interface design; (iii) user state monitoring and structural integrity of the preliminary design. If issues
and personalized regulation; and (iv) personalized driving are identified, the engineering agent suggests modifications
strategy recommendations. The specific implementation and provides feedback to the designer agent. The UX
details for each aspect are elaborated below.
analyst agent assesses the UX aspect of the design based on
3.2.1. Personalized product design the user’s HDT, using simulation and emulation techniques
to model the user’s actions within the cockpit and assess
Utilizing generative model-based MAS and HDT comfort and convenience. Based on feedback from the
technologies to assist in personalized cockpit product UX analyst agent, the designer agent and the engineer
design can significantly enhance design efficiency and agent collaborate to optimize the design. Throughout this
user satisfaction. First, a detailed HDT model, including process, the user agent continuously provides feedback
physical model, psychological, and biological dimensions, to ensure that the design remains aligned with the user’s
is created for each user. This model integrates the needs.
user’s body parameters, preferences, and mental status,
derived from computer vision technology, collected After several iterations, the final cockpit design is
50
heterogeneous data, and user-provided information. generated. Through VR technology, users can experience
During the automated design process, the MAS comprises the final design in an immersive environment and provide
a user agent, designer agent, engineer agent, and UX final feedback. The final design is then validated by the
analyst agent. The user agent leverages the HDT model to engineer’s agent before moving into the production
simulate the user’s actual experience in the virtual cockpit, phase. This approach achieves highly personalized design
providing detailed design requirements and personal through multi-agent collaboration and HDT technologies,

Figure 2. The main functions of generative model-based multi-agent systems and human digital twin based personalized design framework in virtual space

Volume 1 Issue 3 (2024) 9 doi: 10.36922/ijamd.4220