Modeling and Control of a Pesticide Spraying Drone Using the Model Predictive Controller

Abstract

One of the primary sources of income in Ethiopia's economy is Agriculture. Over 80% of the people depend upon the agriculture fields. Agriculture suffers significant losses due to diseases spread by pests and insects, which reduce crop productivity. In order to improve crop quality, pesticides and fertilizers are used to kill insects and pests. However, almost all agricultural fields in Africa, particularly in Ethiopia, use humans to apply pesticides and fertilizers to the entire crop field, resulting in pesticide waste and adverse effects on human health. Unmanned Aerial Vehicles (UAVs) are becoming increasingly popular as agricultural research topics. The purpose of this research is to control the attitude and position of an autonomous pesticide spraying quadcopter drone using a model predictive controller (MPC). Localization, mapping, and trajectory following of a quadrotor are achieved using data from a camera, IMU (Inertial Measurement Unit), GPS (Global Positioning System), and range finder in a disturbance-based environment such as wind and obstacles. In this study, the quadcopter model is modeled using the newton Euler formulation of motion to implement this controller. The Kalman filter (KF) is used to estimate the states of the quadcopter and the variation in quadcopter mass due to spraying. The controller was implemented in the quadcopter using MATLAB to test the feasibility of the proposed control method. The simulation shows that the proposed controller adapts to both setpoint and trajectory tracking. The PX4 firmware includes a Proportional Integral Derivative (PID) controller. MPC, on the other hand, has never been used on commercial drones because it requires a mathematical model of the specific quadcopter to be used. Thus, the purpose of this thesis is to assess the practicability of the MPC control scheme for quadcopters with the modified PX4 firmware to control agricultural drones. The MATLAB/Simulink simulation shows that the proposed controller rejects the disturbance and state noise based on estimator from Kalman filter to the unknown system noise of quadcopter and it performs effectively at the reference of altitude, position and attitude.