Variable-load unmanned aerial vehicles (UAVs) are key tools in smart agriculture, particularly playing a vital role in the prevention and control of crop pests and diseases. Equipped with pesticide spraying equipment, these quadrotor UAVs offer advantages such as high operating speed, low risk of chemical drift, and improved crop coverage, making them widely used in agricultural plant protection. However, during spraying operations, the pesticide liquid gradually diminishes over time, leading to continuous changes in the UAV's overall mass, center of gravity position, and moment of inertia. These time-varying characteristics increase the dynamic complexity of the UAV system, imposing higher requirements for precise trajectory tracking control and attitude response. Current research mostly focuses on modeling under constant-mass assumptions or concentrates on solid suspension systems and abrupt mass changes, often overlooking the continuous dynamic variations caused by the slow loss of liquid payloads in agricultural spraying scenarios. So, how to effectively address the changes in UAV dynamic parameters resulting from such slow liquid loss, ensuring precise trajectory tracking and stable flight during plant protection operations?
To tackle this problem, Dr. Shuting Xu and her team from the School of Technology, Beijing Forestry University, proposed a comprehensive solution. The research first established a time-varying multibody dynamic model for the variable-load UAV, dividing the system into a constant-mass frame module and a time-varying mass pesticide tank module. To accurately describe the impact of liquid loss in the pesticide tank on the UAV's center of gravity and inertia distribution, computational fluid dynamics (CFD) methods were employed. Three-dimensional transient simulations of gas-liquid two-phase flow inside the pesticide tank were conducted using ANSYS Fluent software, considering changes in mass, center of gravity, and moment of inertia during liquid loss. Time-varying functions for these parameters were obtained through curve fitting. On this basis, the frame and pesticide tank models were integrated to form a complete time-varying multibody dynamic model, providing an accurate dynamic foundation for subsequent control design. The related research has been published in Frontiers of Agricultural Science and Engineering (DOI: 10.15302/J-FASE-2025662).
Based on this model, a disturbance-rejection trajectory tracking control system based on PD sliding mode control was designed. The system adopts an inner-outer loop structure, with the inner loop serving as the attitude controller and the outer loop as the trajectory controller. To enhance system robustness and reduce chattering, the reaching law of sliding mode control was improved by replacing the discontinuous sign function with a continuous hyperbolic tangent function, ensuring rapid response to control errors while maintaining stability. Through Lyapunov stability analysis, it was verified that the controller can achieve asymptotic stability of the closed-loop system.
Simulation experiments demonstrated that the controller achieved precise trajectory tracking, with position tracking standard deviations of 0.0507 m, 0.161 m, and 0.0002 m in the horizontal, lateral, and vertical directions, respectively. In terms of attitude control, the roll, pitch, and yaw axes exhibited small transient errors and rapid convergence. Compared with traditional PID control and conventional sliding mode control, this method showed superior performance in trajectory tracking accuracy, dynamic response, and robustness. Flight experiments were conducted in a wheat field in Hebei Province, where the UAV carried a full pesticide tank and flew stably at an altitude of 4 meters, completing the spraying task along the predetermined path. The results indicated high trajectory tracking accuracy in straight-line segments, with deviations of approximately 0.2–0.3 m in turning segments, which could be restored to the predetermined trajectory within 5–8 s, meeting the accuracy requirements for agricultural plant protection operations.
This research provides theoretical and technical support for solving the dynamic control problems of variable-load UAVs in agricultural plant protection, contributing to the improvement of precision and stability in intelligent plant protection. In the future, the team will further investigate directions such as fault-tolerant control, anti-liquid sloshing, and robust algorithms under wind field disturbances, promoting the wider application of variable-load UAVs in agriculture.