HANNOVER MESSE 2019, 01 - 05 April
switch to:
Machine Learning

AI constructs an ultra-aerodynamic bike

To crack the previous record, a French university team wants to send a bike into the race, which has been aerodynamically optimized by artificial intelligence.

22 Jul. 2018
AI constructs an ultra-aerodynamic bike (Photo: EPFL)

The current speed record for a bike on a level track is 133.78 km/h, set in 2012 by a Dutch team in the World Human Powered Speed Challenge . But it could be broken next September. A team from the French university IUT Annecy wants to use AI software to design an aerodynamically optimized bike specially for the competition.

The software is provided by NeuralConcept, a spin-off of the Swiss École polytechnique fédérale de Lausanne (EPFL). After entering the specifications for the length and width of the bike and the space required for the drive and wheels, it calculates the most suitable shape. The program is based on a Convolutional Neural Network , which has been trained with data on the aerodynamic properties of various forms. It tries out different concepts in quick succession and finally presents the optimal construction. According to NeuralConcept, the results are about on average 5 to 20% more aerodynamic than constructions produced with conventional methods. In addition, the software needed only a few minutes for the calculations.