FEUPCAR 2.0: Overview of a Low-cost Robot for Autonomous Driving Competitions

Abstract

Robotic autonomous driving is a very complex task that tries to replicate the human behavior when performing such task. This paper presents a high-level overview of an architecture applicable for small-scale vehicles in autonomous driving competitions. A special emphasis is given on the sensory and navigation sub-systems since they are the most prominent intelligent decision layers. The former is almost entirely based on computer vision, processing the raw image content from two ordinary and inexpensive "web-cams", conveying further relevant information to the later, on a real-time basis. The proposed architecture was implemented with minimal interventions to an Ackermann-like vehicle which was originally designed for recreational purposes, serving as a "proof of concept" of the developed system. Results show that a low-cost, scalable and modular system can be easily integrated on regular small-scale vehicles obtaining exciting results at a minimal cost.