Real-Time Evaluation of Inner-Loop Controllers with NMPC-Based Outer-Loop for UAV

This paper investigates the application of control algorithms for Unmanned Aerial Vehicles (UAVs), focusing on the real-time integration of different inner-loop controllers. A Nonlinear Model Predictive Controller (NMPC) is considered as an outer-loop position controller, while the inner-loop controllers considered are a Geometric controller (with and without L1 augmentation), a Model Reference Adaptive Controller (MRAC), and a Sliding Mode controller. The study focuses on comparing the trajectory-tracking performance of the different inner-loop controllers and evaluating the computational efficiency of the NMPC. These controllers are implemented as ROS 2 nodes to evaluate their suitability for hardware deployment. A performance evaluation of the inner-loop controllers is conducted through Gazebo simulations, evaluating the Root Mean Square Error (RMSE) with respect to a reference trajectory as a merit parameter for tracking accuracy. Additionally, the computational demands of the NMPC are analyzed to assess its real-time feasibility on a hardware target, specifically an NVIDIA Jetson platform, with an evaluation of the computing times for both the outer loop and inner loop controllers.