A new generative AI tool that augments the vehicle design process in multiple ways has been built by experts at the Toyota Research Institute.
Designers could already leverage publicly available text-to-image generative artificial intelligence tools as an early step in the creative process. With TRI’s new tool, initial design sketches and engineering constraints can be added into this process, cutting down the iterations needed to reconcile design and engineering considerations.
“Generative AI tools are often used as inspiration for designers, but cannot handle the complex engineering and safety considerations that go into actual car design. This technique combines Toyota’s traditional engineering strengths with the state-of-the-art capabilities of modern generative AI,” explained Avinash Balachandran, director of TRI’s human interactive driving (HID) division.
Aerodynamic constraints, such as drag, and chassis dimensions, such as ride height and cabin dimensions, can be implicitly integrated into the generative AI process. TRI has tied principles from optimization theory, used extensively for CAE, to text-to-image-based generative AI. The resulting algorithm enables the designer to optimize engineering constraints while maintaining their text-based stylistic prompts to the generative AI process.
As an example, a designer can request, via text prompt, a suite of designs based on an initial prototype sketch with specific stylistic properties such as ‘sleek’, ‘SUV-like’, and ‘modern’, while also optimizing a quantitative performance metric.
“TRI is harnessing the creative power of AI to amplify automobile designers and engineers,” added Charlene Wu, senior director of TRI’s human-centered AI division, whose team collaborated with the human interactive driving team on this project.
By bringing engineering constraints into the design process, the tool could also help Toyota design its EVs more quickly and efficiently.
For more technical details on TRI’s new technique, refer to the following two papers:
Interpreting and Improving Diffusion Models Using the Euclidean Distance Function, F. Permenter, C. Yuan, 2023.
Drag-guided diffusion models for vehicle image generation, N. Arechiga, F. Permenter, B. Song, C. Yuan, 2023.
