Study and Experimentation of Control Policies to Dynamically Maintain Micro-UAV Flight Stability

Unmanned Aerial Vehicles (UAVs or drones, for short) are receiving increasing attention recently, due to the many applications they might be used for, ranging from territory surveillance to delivery of goods. The case of drones autonomously flying according to a pre-defined route is clearly the most interesting one for the deployment of such applications. Yet, autonomic flight requires that a drone is able to maintain stability – in terms of a target attitude – in spite of external disturbances (e.g. gusts of wind), with no human intervention. This is traditionally done with a PID (proportional integral derivative) controller, which takes as input the deviations from the target, and supplies in output the indication of how to act on the drone engines so as to restore the proper attitude. Two orders of problems must be dealt with when deploying a PID controller. First, the parameters weighing the input components must be properly tuned in order to guarantee attitude restoring, and to avoid oscillations of the system. Second, the PID output must be translated into commands to the drone engines so as to achieve the desired behavior. The former aspect can be solved either by hand, or with a number of automatic methods proposed in the literature (e.g. [1, 2]). Those methods involve complex mathematical models and are considerably time-consuming. As far as the latter aspect is concerned, there are hardware constraints that impede to abruptly change speed or spin of rotors and propellers, which otherwise may be damaged. In this paper, we introduce a mathematical model simplified yet effective in yielding appropriate parameters to implement an accurate controller, without the need of a complex preliminary calibration of the mechanical system. We describe our policies to apply controller indications to the drone hardware. We validate both model and policies through experiments.