We propose an applied machine learning course that teaches students with no machine learning background how to train and use deep learning models for deploying aerial drones (multi-copters). Our unique, hands-on curriculum gives students insight into the algorithms that power autonomous systems as well as the hardware technology on which they execute. Students learn how to integrate Python code with serial communications for streaming sensors and imagery to deep learning models. Students use OpenCV, Keras, and TensorFlow to learn about computer vision and deep learning. The final project (see Figure 1) provides the opportunity for students to plan and develop an end-to-end, fully autonomous, self-contained product (i.e. all systems physically residing on the drone itself) that is integrated with heavy-payload drones and computer vision in a scenario centered around an outdoor search and rescue mission. With no human in the loop, students deploy drones in search of a missing person. The drone locates and identifies the individual, delivers a care package to their location, and then reports the individual’s geolocation to ground rescuers before returning home. The novel helper code and solutions are built in-house using Python and open technologies. Results from a pilot offering in the spring of 2021 indicate that our approach is effective and engaging for computer and cyber science students who have previously taken a basic artificial intelligence course and who have 1-2 years of programming experience. This paper details the design, focus, and methodology behind our Autonomous Systems Integration curriculum as well as the challenges we encountered during its debut. © © 2022 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved.
Author Keywords
Index Keywords
