Artificial intelligence (AI) continues to reshape various sectors, with notable advancements emerging within the automotive industry. Creative Bloq reports that the potential of AI has particularly caught the attention of car manufacturers, especially concerning the design and development of electric vehicles (EVs). The costs and time involved in creating new car models are significant, primarily due to the numerous iterations required to refine designs and prototypes.

Electric vehicle design has garnered mixed reactions over the years, as evidenced by public opinions on various concepts, from the Tesla Cybertruck's polarising aesthetics to the innovative Jaguar Type 00. Recent challenges faced by the Cybertruck, including its performance in snowy conditions, highlight the complexities involved in EV design. In response to these challenges, researchers at the Massachusetts Institute of Technology (MIT) believe that AI can streamline the vehicle design process. They have developed an open-source database known as DrivAerNet++.

The DrivAerNet++ database, which was built upon 39 terabytes of data, leveraged an impressive 3 million hours of central processing unit time using the MIT SuperCloud. It comprises over 8,000 3D models based on existing car designs, created through an algorithm that assessed 26 different parameters. These parameters include vehicle length, underbody features, windshield slope, tread, and wheel shapes corresponding to various baseline models.

In an additional step, the research team employed an algorithm to ensure that newly generated designs were unique and not mere replicas of existing vehicles. The resulting 3D designs were translated into accessible formats, including mesh and point cloud representations, along with a comprehensive list of dimensions and specifications. To evaluate the aerodynamic performance of each design, fluid dynamics simulations were conducted, analysing how air would circulate around every generated model.

The primary objective of creating this expansive dataset is to facilitate the training of AI models, which could identify optimal feature combinations, balancing aerodynamic form and eco-friendly engine efficiency. This advancement aims to significantly reduce research and development costs while accelerating the car design process.

Faez Ahmed, an assistant professor of mechanical engineering at MIT, elaborates on the implications of this research, stating, "The forward process is so expensive that manufacturers can only tweak a car a little bit from one version to the next. But if you have larger datasets where you know the performance of each design, now you can train machine-learning models to iterate fast so you are more likely to get a better design."

The team presented their findings at the NeurIPS conference held in Vancouver in December, marking a significant step towards utilising AI for more innovative and economically viable car design solutions in the future. As the automotive industry continues to evolve with technology, the intersection of AI and engineering may well pave the way for advancements that could redefine how vehicles are conceptualized and manufactured.

Source: Noah Wire Services