Artificial intelligence might soon eliminate one of the biggest bottlenecks in the car-design process.
Toyota Research Institute executives said Tuesday they’ve developed a generative AI technique that ensures early design sketches incorporate engineering parameters.
Using the new technique could curtail the number of iterations needed and shorten design cycles, said Avinash Balachandran, director of TRI’s Human Interactive Driving division, whose team developed the technology.
“When a designer tries something, it typically goes to engineering, and they say, ‘This is not possible and that is not possible,’ and it goes back to the designer,” he said. “That’s one reason design iteration takes a long .
Now designers can impose restraints, such as asking the AI to meet specific requirements on drag.
That’s an immediate focus for the company, given that aerodynamic gains would increase the range of battery electric vehicles.
In the future, they intend to explore using AI to incorporate handling characteristics based on wheelbase and structural integrity standards and dynamic performance measures.
The development of high-end battery electric vehicles has led to the transformation of vehicles’ computing processes – using centralized or zonal architectures that reduce the number of electronic control units (ECUs) and wiring in the vehicle.
Auto meets AI
Toyota’s work builds on existing text-to-image techniques, in which designers can start with a sketch and prompt the AI tool to provide variations of the image — for instance, by asking it to make it look sleeker or more like an SUV.
Elsewhere in the realm of vehicle design, Nvidia has relied on its AI and simulation technologies to create its proprietary Omniverse — a platform that allows for 3-D workflows.
Companies like BMW are using Omniverse to track updates in vehicle design which can immediately ripple through factory planning.
To support, not replace, humans
ChatGPT and generative AI have become recent buzzwords, but TRI has been developing its design techniques for about 18 months. While AI’s world-changing potential remains the subject of widespread fascination and fear, Toyota wanted to address a specific, practical problem.
There’s no metric yet that quantifies how Toyota’s AI technique might shorten the design and iterative process, but Toyota hopes “designers will explore an idea much faster,” Balachandran said, including the portion where “they’re looking out the window searching for inspiration.”
The company says it is not looking to replace human designers but to amplify their work. Supporting humans with technology is a pillar of TRI’s work across five research areas, which include robotics, human-centered AI, human interactive driving, machine learning and energy.
Lessons from self-driving technology development underscored the need for machines to collaborate with humans, not replace them, Balachandran said.
“There are insights an expert can add that AI doesn’t capture,” he said. “So we’re looking for the best blend of both.”
