Comparative Study of Particle Swarm Optimization and Genetic Algorithm for Automated LQR Gain Tuning in Quadrotor Trajectory Tracking

Abstract

Drone applications in time-critical scenarios such as urgent medical deliveries and post-disaster area mapping require robust flight controllers that ensure effective performance while minimising actuator effort. The Linear Quadratic Regulator (LQR) offers high performance and robustness; however, the absence of a systematic method for selecting its weighting matrices means performance is not guaranteed and depends on the designer’s expertise. This paper addresses that gap by evaluating and comparing Particle Swarm Optimization (PSO) and the Genetic Algorithm (GA) for automated determination of optimal LQR weighting matrices for quadrotor trajectory tracking. Performance is assessed via trajectory feasibility, position error at each waypoint, and convergence speed. PSO consistently achieved a lower cost (72.83) than GA (74.94 best-case) and converged in under one second, versus 77–92 s for GA. PSO also produced smoother inter-waypoint transitions, making it the more suitable algorithm for this application.