Fast Trajectory Tracking for Commercial Off-the-Shelf UAS Using Model Predictive Control: Real-Time Implementation and Performance Analysis

Following swift changes in the waypoints of a trajectory is crucial for unmanned aerial systems (UASs). This capability allows UAS to quickly adapt to changing conditions and execute complex maneuvers with high accuracy, enhancing their effectiveness in various applications. Implementing trajectory tracking in real-time is challenging, even more so if the UAS is a Commercial off-the-shelf (COTS) platform. Model Predictive Control (MPC) is a potential solution for this scenario, however, the dynamic model of UAS is essential for proper functionality. Obtaining an exact model of a COTS UAS is often difficult, especially when dealing with platforms executing internal controllers on an embedded autopilot. To overcome this obstacle, our study implements a model identification process for the UAS, through processing input/output data pair collected during experimental flights. Using the identified model, the MPC optimizes the error between the system's current state and the desired reference trajectory over a finite horizon and calculates the appropriate control command while considering the system velocity constraints. The functionality of the tracking mechanism is demonstrated under two real-time trajectory tracking scenarios, a circle, and a rhodonea curve, executed at translational velocities close to the physical constraints of the UAS. The experimental evaluation of the proposed approach, focusing on accuracy and control performance, demonstrates the potential of the proposed solution for COTS UAS platforms.