Technology is being used in areas a person will not think much about. Machine Learning is a trend and people are trying to find its application in various locations. A Swiss startup is trying to use Machine Learning to make the world’s fastest human-powered bike.
For now, the world record for the fastest speed ever achieved in riding a human-powered bike on a straight road is 133.78 km/h. It was set by Sebastiaan Bowier, member of a Dutch team which set the record in 2012. They achieved 83 mph on a bike which was driven on a straight road. But, as it is said, records are not kept for a longer time. They are broken now or later.
Reportedly, the record is soon to be broken by a Swiss Startup. They have planned to beat the speed time by using their very own vehicle. The vehicle is built using the help of machine learning. The startup has developed an algorithm using the machine learning which will help streamline their new human-powered bike. The new design will cut the air-resistance giving them extra speed to beat the record easily.
The Fastest Human Powered Bike
The bike looks just like a small rocket which is bounded to land. The design seems a little weird but if someone wants to set a record, every bit of it counts. No one knows if using this design might make them achieve the top speed beating the previous record by 1/100th of a mile. Hence, it becomes incredibly important to maintain the shape and design of the bike.
The algorithm which is based on the Machine Learning is developed at Ecole Polytechnique Federale De Lausanne’s Computer Vision Laboratory. The algorithm is trained to work on 3D shapes and to improvise the aerodynamic qualities. Pierre Baque, one of the researchers said that “It will work on the laws of physics and improve itself. The algorithm developed takes an image as input for the system. Images are well made hence it makes the work easy. But, most of the time the input is in the form of a mesh. Mesh is typically a graph with lot many points and are difficult to handle”.
He added that “In spite of all this, his team of researchers were able to develop a neural network which will sort through all of these. And, in the end, it will provide the best aerodynamic design which can be made out of it. The algorithm is efficient that it gives around 5% to 20% more efficient design than the conventional methods. Moreover, the algorithm can be used in situations when there are no conventional methods available. Hence, it increases the availability and predicts better design”.
In the International Conference on Machine Learning in Stockholm, Peter Barque presented the new Neural Network. It was presented as a product of the different company and Peter is said to be its CEO. Further, it was stated that the technology will be applied to test it on a vehicle which will be developed by September. Since it is a human-powered machine, it has to be made according to the humans. The model will be presented at the World Human Powered Speed Challenge which will be held in Nevada.