Control System Design and Management for Autonomous Vineyard Harvesting Robot

The master thesis details the design and development of an autonomous control system for a robotic manipulator tailored to grape harvesting applications. The proposed system enables the robotic manipulator to autonomously detect, approach, grasp, cut, and deposit grapes in a designated collection zone through a fully automated operational sequence. This manipulator, offering six degrees of freedom powered by stepper motors, executes precise movements managed by a low-level control architecture centred on an STM32NUCLEO-F401RE microcontroller interfaced with EVLPOWERSTEP01 motor drivers, under the coordination of an NVIDIA Jetson Nano Orin. The system’s SPI-based communication between the NUCLEO microcontroller and motor drivers allows for seamless real-time motor command execution and parameter monitoring. The Jetson Nano Orin leverages a vision-based detection system, incorporating a webcam to capture spatial data, enabling it to drive real-time decision-making for harvesting via multithreaded control algorithms. These algorithms, structured as a state machine, coordinate key steps in the harvesting cycle—grape detection, alignment, grasping, cutting, and release. Performance assessments demonstrate high positioning accuracy and efficient operation, optimized for use within constrained vineyard rows due to the manipulator’s limited range of motion. To further enhance operational reliability, a manual control mode was integrated, accessible through either joystick or keyboard inputs. This feature, developed in Python and integrated within the system’s state machine, permits direct operator control over robotic functions, offering flexibility and safeguarding against potential system errors. This thesis establishes the viability of autonomous robotic grape harvesting and underscores its potential for refinement and broader application in viticulture.