Trajectory tracking and state estimation for quadcopter UAVs using particle filters in an inner-outer loops controller
Vitorino Biazi, Jan Nedoma, Carlos Marques · Expert Systems with Applications · 2025
Unmanned Aerial Vehicles (UAVs), such as quadcopters, have become widely used in high-risk applications, such as firefighting and rescue in earthquake regions, requiring reliability and resilience to noise and possible sensor failures. Traditional state estimators like the Extended Kalman Filter (EKF) handle Gaussian noise effectively but struggle with nonlinearities and non-Gaussian noise. In contrast, Particle Filters (PF) offer greater robustness and adaptability to such conditions. This paper presents a state estimation approach for quadcopters based on PF integrated with an inner-outer loops controller to achieve precise positioning and trajectory tracking. The study aims to validate the efficiency of the proposed PF-based algorithm for UAV state estimation, even amidst noisy sensor data and limited measurements. Computational simulations using the Julia programming language were conducted for a specific quadcopter model and two different tasks were tested: spatial positioning and trajectory tracking. The obtained results demonstrate that the PF algorithm effectively estimated the drone’s state, allowing the control system to guide the UAV to the desired position or trajectory with high repeatability and accuracy. For the positioning task, the mean computational processing time and settling time interval are 0.108 s and 1.6–2.6 s for 10 particles, 0.155 s and 1.8–15.2 s for 100 particles and 0.947 s and 2.0–35.6 s for 1000 particles, respectively. The results indicate that increasing the number of filter particles reduces initial variability in state estimates. However, it also increases posterior variability, computational costs, and convergence times, highlighting the need to balance particle count for optimal system performance.