Browsing by Author "Elisabeth Andarge"

Now showing 1 - 1 of 1
  • No Thumbnail Available
    Item
    Trajectory Tracking Control of Quadcopter using Fuzzy Super Twisting SMC with PID Surface for Wheat Yield Estimation
    (Addis Ababa University, 2023-11) Elisabeth Andarge; Elisabeth Andarge (PhD)
    Accurate yield estimation during the heading stage of wheat production is pivotal for efficient harvest planning and ensuring food security. However, traditional yield estimation technique is labor-intensive and potentially harmful to the crop. Alternatively, the use of satellite imagery for estimation is hindered by its poor resolution. Another approach involves employing low-altitude quadcopters to capture high-resolution images. Yet, controlling flight of these quadcopters presents challenges due to their nonlinear and underactuated characteristics. This thesis aims to address these challenges by designing a flight controller capable of controlling the quadcopter to track the desired trajectory using Fuzzy super twisting sliding mode controller with PID surface to capture images and estimate wheat yield. The process begins with modeling the quadcopter using the Newton Euler method, followed by designing a flight controller. This controller is divided into inner and outer loops capable of automatically adjusting its parameters, and the system is then simulated in MATLAB/Simulink. Wheat head images acquired from the quadcopter are trained using transfer learning in YOLOv8. Parameters such as kernel weight, the number of kernels per head, and sampled area are extracted from the field. The controller’s performance is rigorously assessed through various trajectory and disturbance scenarios, comparing it with sliding mode controller both with and without parameter variation. The results demonstrate the controller’s efficiency in guiding the quadcopter along predefined trajectories, robustly rejecting disturbances, and effectively handling parameter variations. In the realm of image processing, the system exhibited notable advancements in both training and validation accuracy. In summary, the proposed controller enhances robustness, the capacity to handle parameter variations, disturbance rejection, chattering minimization, and controller effort reduction. Utilizing the trained weight parameters, the system can accurately detect and count wheat heads, ultimately providing an estimation of wheat yield.