Deep Learning Applied to Autonomous Vehicles - Current Applications and Perspectives
Machine learning and especially deep learning are useful tools for dealing with autonomous vehicle problems that are particularly hard to model, such as obstacle identifications and intentional predictions. In a potentially complete migration to fully automated roads, they are also a solution to deal with the aspects of shared roads between fully autonomous and communicating vehicles and traditional vehicles where driving is not predictable. In spite of attractive aspects of these solutions, their integration in autonomous vehicles must also deal with computing time issues and correct training. Finally, their applications once problems solved are numerous to the point of being able to envisage a system mainly composed of neural networks.
Steve Pechberti is currently a researcher at VEDECOM. He completed his Ph.D. about 'electromagnetic sensors modeling applied to autonomous vehicle' at Université d'Evry in 2013 after his engineer studies at ISTIL Lyon. His research interests lie in the area of modeling, artificial intelligence and computer sciences, ranging from theory to design. Since 10 years, he works on various thematics in autonomous vehicles and has designed several full-stack embedded architectures; in recent years, he has focused on artificial intelligence applied to autonomous vehicles. In his spare time, he conceives connected domotic solutions and spends time in nature parks.