Flight Controller Design for a Helicopter Based Unmanned Aerial Vehicle (UAV) Using Differential Evolution-Based PID Tuning

Abstract

Helicopters possess the ability to perform a range of maneuvers, including vertical takeoff, hovering, forward or sideways movement, spinning, and turning. Helicopter UAVs are highly advantageous for both military and civilian applications due to their exceptional agility, particularly in hazardous environments. It is critical to evaluate the control system's performance and safety through simulation before using it on actual flights. This helps to reduce risks and minimize the need for extensive flight testing using complex aeronautical systems. The objective of this study is to develop an autonomous flight control system (AFCS) specifically designed for a helicopter-based unmanned aerial vehicle (UAV). This system will control and regulate the helicopter's altitude, attitude, velocity, and heading. The purpose of the study is to determine the optimal gains for a PID controller using the Differential Evolution methodology. Differential evolution is a heuristic method that operates on a population of vectors. To minimize cost functions, it replaces the existing vectors with trial vectors derived from mutant vectors. The optimization process consists of four distinct phases: initialization, mutation, crossover, and selection. The method is a simple and efficient search approach for finding the global optimum. It employs population model functions with real-valued arguments, as well as those defined on completely ordered spaces. The control algorithm underwent testing and simulation using the X-Plane 9 flying simulator and LabVIEW software. A sequence of simulated flight tests validates the performance and effectiveness of the proposed controller. Therefore, it can be inferred that the controller gain obtained by the Differential Evolution (DE) approach in the AFCS design is suitable for hovering and circular trajectory tracking. This is achieved by adjusting the PID gain values based on the Q matrix variations in a multi-objective cost function. The accuracy of position tracking is assessed using the mean absolute error. Both trajectories were measured, yielding a low error and indicating good accuracy.