MIT engineers have developed the largest open-source dataset of car designs, including aerodynamic data, which could accelerate the creation of eco-friendly and electric vehicles. Named DrivAerNet++, the dataset features over 8,000 car designs generated from today’s most common vehicle types. Each design is available in 3D format and includes detailed aerodynamics simulations, providing crucial insights into airflow around the vehicle. The data is accessible in multiple formats, including mesh, point cloud, and parametric lists, making it compatible with various AI models.
Quick analysis of vast amounts of data
MIT researchers believe AI-driven design could revolutionize the automotive industry by quickly analyzing vast amounts of data to generate optimized vehicle concepts. While generative AI tools exist, the lack of comprehensive, structured data has been a major obstacle—one that DrivAerNet++ aims to overcome. Aerodynamics play a critical role in vehicle efficiency, particularly for EV range.
The dataset allows researchers to train AI models to recognize ideal aerodynamic configurations and instantly generate optimized designs. Conversely, engineers can input a specific car design into an AI model trained on DrivAerNet++ to estimate its aerodynamic properties and predict fuel efficiency or EV range.
(Courtesy: Mohamed Elrefaie)
