This project presents a cost-effective solution for achieving autonomous multi-drone control for cinematography using DJI Tello drones. Through our exploration of multi-drone systems, we developed a fully customized multi-drone interface, composed of five main components: mapping, localization, collision avoidance, movement control, and subject tracking. Using this framework, our method allowed multiple drones to move autonomously from one user-selected location on a map to another along a straight-line path while continuously tracking a human subject through employing usage of computer vision, specifically a Convolutional Neural Network. By using IMU odometry, we eliminated the need for complex and expensive systems such as LiDAR and GPS. Furthermore, the interface featured a custom collision avoidance algorithm, utilizing a drone’s three-dimensional maneuverability to avoid collisions and prevent the drones from obstructing each other’s video feeds. In our tests of the drone system, both drones were able to move simultaneously to different desired locations on our map based on user-defined start and end locations, with real-time position estimation averaging an 80. 4% accuracy. Furthermore, the subject tracking algorithm kept the subject in the frame with an average accuracy of 87. 8%. Additionally, the drones reached destinations with an accuracy of 92. 2%. The largest factor affecting the system’s performance was wind speed. Overall, with a total system cost of 262, we created an interface which allows users to control multiple autonomous drones at once, testing the effectiveness of this interface on two drones in a real-life environment. The system enables drones to navigate to any location on a map while continuously tracking a subject for enhanced cinematography and providing an easily scalable and accessible program.
Bai et al. (2025) studied this question.