Low Cost Platform for Teaching AI Self-Driving Cars Topics for Undergraduate Students in Emerging Countries

Abstract

This full paper presents the validation and results of a low cost scaled car platform into a project-based course in order to teach AI self-driving cars topics for undergraduate programs in Universities. This is an elective course of the Mechatronics program at Pontificia Universidad Catolica del Peru (PUCP) whose second edition of the course was developed during January through March 2020. The main objective of this article is to present the results of the second edition of the project-based course, which details the integration of a low cost robotic platform with an embedded board used to execute computer vision and AI algorithms. Using a robotic platform allowed the students to focus on the application of the algorithms in a real scenario and learn from experience instead of using only simulation platforms. The proposed course aims to introduce the students in self-driving cars topics, and apply the theoretical concepts to develop an autonomous car using the robotic platform. The topics of the course are structured in five categories including Automotive Design Concepts, Localization and Navigation, Computer Vision Techniques, Artificial Intelligent Techniques and Simulation Environment; and is divided into fourteen theoretical lectures and five practical laboratories. The project-based course is aligned with four Students Outcomes from ABET accreditation entity for undergraduate programs in order to reinforce their abilities to work as a team, self-learning, hands-on experience, develop prototypes, testing in real scenarios, and learn basic scientific writing and presentation skills. The results of the second edition of the course show that the students enrolled were able to accomplish the development of a self-driving car capable of completing a lap on a racetrack autonomously only using image processing and AI algorithms. In comparison with the first edition of the course, the inclusion of a scaled car as a base for the project avoided mechanical problems with the chassis and allowed the students to focus on the sensors integration and algorithms programming.