École Nationale Supérieure de
Techniques Avancées

Abstract

Today, artificial intelligence is becoming an indispensable tool in all fields. The methods developed and recent developments in machine learning are working well and will gradually replace “traditional” approaches. Everyone is eager to automate everything so that the human being is only a supervisor. We’ve been working on a topic of prediction and learning about the traversability of a robot. First, using the “robot operating system”, we created a real-world image data set. This helped me control the recording and modification of this base according to project needs. Second, this database used to train and test different neural networks. Finally, I launched the “hyperband” algorithm to optimize the values of the hyper-parameters in order to obtain better results.