Page 11 - IJAMD-1-3

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

overstated. Devices such as smartphones, wearables, and potential that surpasses traditional design methodologies.
Internet of Things (IoT)-enabled gadgets are prevalent Generative models, exemplified by Generative Pre-trained
in our daily lives, facilitating seamless connectivity and Transformer 4 (GPT-4) released by OpenAI, have initiated
access to information. For smart vehicles, this connectivity a paradigm shift in the field of artificial intelligence (AI).
is leveraged to create a more integrated and responsive These models enable users to intuitively generate a vast array
6
driving environment, allowing the cockpit to adapt to the of high-quality, multimodal content in a short time span,
driver’s needs and preferences in real time. Generative encompassing text, images, video, and interactive 3D content
7
artificial intelligence (GAI) has experienced unprecedented (e.g., avatars, 3D models, and 3D environments ) based on
growth; with applications like Chat Generative Pre-Trained users’ instructions. MAS are composed of autonomous
8
Transformer (ChatGPT) emerging as the fastest-adopted entities known as agents, which possess inherent learning
consumer software in history and can further amplify the and decision-making abilities and can collaborate internally
3
9
capabilities of smart vehicle cockpits. Technologies such as to address complex design tasks. Moreover, generative
generative models and deep learning algorithms facilitate models present a promising way for augmenting the
the processing of vast amounts of data from both the vehicle capabilities of agents within MAS.
and the driver. This enables an advanced decision-making Meanwhile, HDT refers to the digital representation of
process that can lead to the development of personalized human beings in the physical world, which relies on the
10
design, adaptive user interfaces, personalized driving continuous collection of data through wearables and sensors
strategies, and real-time monitoring and regulation of that capture crucial health metrics, physical activities,
driver states. personal preferences, and environmental interactions. This
Meanwhile, users have higher expectations for extensive data can reflect both physiological features and
personalized and humanized cockpit design and anticipate intrinsic cognitive characteristics of humans in a virtual
that the cockpit can provide tailored products and services space, thereby prioritizing physical and mental health
11
according to their own preferences and real-time status, in smart product-service systems. Furthermore, HDT
so exploring the personalized design of the smart vehicle can perform activities on behalf of humans in virtual
cockpits is of great significance. Personalized design is environments, overcoming the physical constraints of the
4
crucial for addressing the diverse needs of users, enhancing real world and broadening the scope of human activities. 10
the user experience (UX), fostering product innovation, Therefore, this study proposes CockpitGemini, a novel
and maintaining enterprise competitiveness. The design framework that integrates generative models, MAS,
5
personalized design of smart vehicle cockpits offers drivers and HDT technologies, enabling highly personalized
unique products, interactive interfaces, environment design for smart vehicle cockpits. The framework can
adjustments, and tailored driving strategies, thereby efficiently provide drivers with unique product designs,
enhancing driving safety, comfort, and the overall UX. This interactive interfaces, environment adjustments, and
approach not only satisfies the requirements of various driving strategies based on their preferences and real-time
clients but also advances the development of smart vehicle status, thereby realizing an overall personalized experience.
cockpits toward greater user-friendliness and intelligence. This integration enhances the personalization of services
Currently, although some luxury vehicles have started to and products, ultimately leading to a more user-centric
incorporate personalization features, the constraints of approach in the vehicle sector. The major innovation of
traditional vehicle architecture prevent most mainstream CockpitGemini can be summarized as follows:
models from achieving true personalized design. Existing (i) This study presents a novel personalized design
solutions typically involve basic parameter adjustments and framework that combines generative model-based
lack a profound understanding of the driver’s state and the MAS with HDT models to enable the efficient delivery
capacity for flexible response. Furthermore, most current of personalized smart cockpit designs and services
research is confined to individual technical aspects, such as based on user preferences and states.
the personalization of human-machine interfaces or driver (ii) The four primary functions achieved by the innovative
assistance systems, without a holistic and comprehensive design framework are demonstrated in personalized
design framework. Integrating advanced AI techniques, product design, personalized interactive interface design,
such as generative models, multi-agent systems (MAS), user state monitoring and personalized regulation, and
and digital twins (DTs), into the personalized design of personalized driving strategy recommendations.
smart vehicle cockpits remains a significant challenge.
(iii) An elaborated case study of personalized vehicle seat
Considering existing constraints, generative model-based design is presented to show the feasibility and usability
MAS and human digital twin (HDT) have demonstrated of the proposed personalized design framework.

Volume 1 Issue 3 (2024) 5 doi: 10.36922/ijamd.4220