Following and Overtaking: A Policy for Autonomous Car Driving

Abstract

In this work, we proposed an approach for autonomous driving based on the concept of following and overtaking. This policy controls the vehicle to follow a car ahead of it, and when getting closer, it appliesan overtaking step to pass the car and later on to getin corporated to the lane again. For this work, we exploit the robustness of Convolutional Neural Networks for object tracking, which is used to track the car ahead of our autonomous vehicle. We use the pixel position of the tracker in combination with measurements from a laser scanner sensor, as input signals in a PID controller, responsible for driving the vehicle autonomously. We have carried out evaluations of our proposed policy in the Gazebo simulator, whose results indicate the feasibility of our approach.