The paper presents CineMPC, a system that autonomously controls a drone for filming purposes. Unlike existing solutions that only control the camera’s position and orientation, CineMPC also includes the camera’s intrinsic parameters like focus, depth-of-field, and zoom. This allows the system to fulfill both the artistic and technical requirements specified by the user. The drone and camera are controlled in a nonlinear Model Predicted Control (MPC) loop, optimizing the trajectory at each time step based on the scene’s current conditions. The system’s perception can track the targets’ position and orientation despite the camera effects. The system has been tested in a photo-realistic simulation and with a real platform, demonstrating its capabilities.
Publication date: 12 Jan 2024
Project Page: https://github.com/ppueyor/CineMPC_ros
Paper: https://arxiv.org/pdf/2401.05272
