Design and implementation of a robotic pressure-sensing hand for an autonomous grape harvesting system

This master’s thesis presents the mechanical and electrical design of a hand-like end-effector integrated into an innovative autonomous grape harvesting system. To interact safely and effectively with delicate, irregularly shaped fruit such as grapes, the end-effector, and its control system were designed to replicate the movements and approach typically employed by human harvesters. A core aspect of this design is the skin-like pressure sensor network embedded within the robotic hand, enabling the end-effector to continuously monitor and regulate the applied force during interactions with the fruit, ensuring optimal handling. This sensor network was intentionally built with scalability in mind, allowing future system iterations to enhance the “resolution” of pressure sensing capabilities, thus facilitating more refined control during harvesting tasks. The development of the robotic hand commenced with the creation of a custom printed circuit board capable of controlling the finger motors, gathering and preprocessing sensor data, and communicating with the system’s central computational unit and other peripherals via a low-energy Bluetooth protocol. Following the development of the PCB, the focus shifted first to writing custom firmware to manage all the needed operations and subsequently to the mechanical redesign of the preexisting hand. One of the primary objectives during this redesign was to increase the range of motion of the fingers, thereby enhancing the flexibility and adaptability of the end-effector. Concurrently, efforts were made to reduce the weight and size of the hand, enabling it to operate more effectively within the complex and often cramped environment of a vineyard. Additionally, custom mounting hardware was integrated to accommodate both the control boards and the skin-like sensor network within the end-effector. Extensive testing and calibrations were conducted to ensure the proper operation of the custom PCBs, with a strong emphasis on verifying Bluetooth communication and ensuring reliable interaction with the sensor network. The final stage of development focused on integrating the hand within the overall system, which comprises a 3-DOF robotic arm where the hand serves as the end-effector, along with a smaller 6-DOF arm specifically designed for cutting grape stems. Once all components were assembled, comprehensive tests were conducted on the entire system within an artificial environment meticulously constructed to replicate, as closely as possible, the conditions found in a real vineyard. This testing phase aimed to verify the overall capabilities of the system in effectively employing a machine vision algorithm to accurately identify grapes, successfully grasp them without causing damage, and efficiently harvest them in a manner that meets modern agricultural standards.