The Experience of Robesafe Team in CARLA Autonomous Driving Challenge

Abstract

The future of the automotive is focused on achieving total autonomous cars in realistic urban environments. To reach it, many researching teams are working with 3D simulators such as V-REP and Gazebo, due to an easy integration with ROS platform. ROS is a middle-ware for robot code development. It allows easy communication between different systems. It is multilanguage, admitting C++ and Python code programming. Those simulators provide precise motion information, but they are designed for smaller environments like robotic arms, providing unrealistic appearance and very slow performance, being unrecommended for real-time systems in rich worlds like urban cities. CARLA simulator provides high detailed environments in realistic urban situations, being able to train and test control and perception algorithms in complex traffic scenarios, very close to real situations. CARLA Autonomous Driving Challenge was launched in Summer 2019, allowing everyone to test their own control techniques under the same traffic scenarios, scoring its performance regarding traffic rules. Robesafe researching group, from Universidad de Alcalá, submitted to this challenge, with the aim of achieving high results and compare our control and perception techniques with others provided by other teams.