Control of an autonomous system for grape harvesting

This thesis presents the design and development of an autonomous robot specifically engineered for grape harvesting. The robot comprises a primary manipulator, a hand-like end effector equipped with pressure sensors, and a cutting arm. The work primarily focuses on the implementation of a positioning algorithm for the cutting arm, as well as the establishment of a Bluetooth Low Energy (BLE) protocol for communication with the motor control boards. The drivers were developed using STM32CubeIDE for programming the STM32WB5MMG microcontroller, while the positioning algorithm is written in Python to be executed on the NVIDIA Jetson Orin Nano, along with the master’s side of the Bluetooth connection through the Bleak library. The initial phase involved the creation of motor drivers for the cutting arm, utilizing DC servo motors based on a linear relationship between output angle and input duty cycle. Key motor parameters were quantified, including angle and duty cycle ranges. The motors are controlled using TIM2 in PWM mode, set to a frequency of 1 MHz. The following phase addressed the implementation of an inverse kinematics algorithm to accurately position the end effector of the cutting arm. Given the confined workspace above the manipulator, a five-joint structure was adopted, allowing the shears to approach the grape stem from various orientations, thereby enhancing operational flexibility. The manipulator incorporates both standard components and custom-designed parts, created in FreeCAD, to accommodate the shears, the actuating motor, and the camera. Initially, the Denavit-Hartenberg parameters were established to define the system's geometry, followed by the implementation of functions to handle matrices and to run the positioning algorithm. This algorithm moves the Tool Center Point (TCP) of the end effector accurately while addressing the orientation of the TCP as a secondary task to ensure a faster convergence, optimizing for the perpendicular alignment of the shears to the grape stem. The final phase of this thesis involved enabling communication between the control boards and the Jetson via Bluetooth. Each board implements the Generic Access Profile (GAP) for advertising and the Generic Attribute Profile (GATT) for data exchange. A singular service is established with three characteristics: a write characteristic for command reception, a read characteristic for sensor data updates, and a notify characteristic for sending messages. The transmission utilizes a broadcast method, whereby the Jetson sends the same command to all BLE devices. Each device decodes the received command to verify its relevance and sends a notification upon execution. Sensor data are continuously updated within the read characteristic and can be accessed as required by the Jetson. The prototype underwent testing in a controlled environment, demonstrating the system's capability to identify and grasp grapes without causing damage.