Road context classifier and route mapping for Autonomous Vehicles

Abstract

Investment in autonomous vehicles has been increasing in recent years, where only developed countries have the necessary infrastructure to integrate autonomous vehicles effectively into their roads. For countries with a lower quality of infrastructure, such as Latin American countries, it becomes a major challenge in terms of technology, algorithms, and investment. The present work implements a context classifier and route mapping system through a convolutional neural network trained with images of different road contexts in Mexico. The system localizes and maps the route using GNSS (Global Navigation Satellite System) and an interactive web user interface, allowing the streets' interactive analysis. The system enables the vehicles to obtain the type of road context expected in the routes it drives. Thus, it provides support and security in integrating autonomous driving functionalities in routes where the road context and the infrastructure allow them.