At Supercomputing 2018 in Dallas, Texas, Swiss-based startup Neural Concept displayed their ultra-aerodynamic bike. The bike was developed using the company’s cloud-based machine learning software that leverages the power of NVIDIA GPUs.
“Our program results in designs that are sometimes 5–20% more aerodynamic than conventional methods. But even more importantly, it can be used in certain situations that conventional methods can’t,” Pierre Baqué, CEO of Neural Concept told EPFL News.
Another benefit of using the machine learning software is that it can compare designs without human bias, Baqué said. “The shapes used in training the program can be very different from the standard shapes for a given object. That gives it a great deal of flexibility,” adds Baqué.
Using NVIDIA TITAN X GPUs and NVIDIA Tesla P100 GPUs the team trained a convolutional neural network to calculate the aerodynamic properties of various forms including drones, bikes, cars and other objects.
The algorithm can calculate the polygon meshes, which are collections of points used to generate 3D shapes, of the objects it’s applied to with great accuracy and efficiency.
Specifically for the bike, the team used AI to boost the performance of its bike. In just a short time the algorithm can calculate the optimal shape of a bike to make it as aerodynamic as possible.
In September the team won the French national record for a bicycle traveling across a flat road. They hope to beat the world record of 133.78 km/h, set in 2012 by a Dutch team at the World Human Powered Speed Challenge which takes place in the Nevada desert.