Abstract
dc:descriptionHuman–robot interaction (HRI) is increasingly recognized as a transformative approach in modern agriculture, with the potential to significantly enhance the efficiency, flexibility, and adaptability of farming practices. As an inherently interdisciplinary field, HRI synthesizes advancements in information and communication technologies (ICTs), robotics, machine learning, and cognitive sciences, enabling the development of innovative solutions to the complex and dynamic nature of agricultural environments. The integration of these technologies facilitates a more holistic and intelligent approach to agricultural automation, addressing both labor-intensive and precision-demanding tasks. A review of the existing literature reveals that HRI applications are predominantly concentrated in high-value crops, where the return on investment justifies the technological adoption. Notably, crops such as walnuts, grapes, and strawberries have emerged as focal points for HRI research and deployment. Within these contexts, modes of interaction that emphasize collaboration and cooperation between human workers and robotic systems, often involving varying degrees of autonomy, have been identified as the most effective. Across a range of implementation scenarios, the collaborative synergy between humans and robots has shown significant promise, yielding improvements in system performance, reducing the physical workload of laborers, and ensuring efficient task execution. However, despite these encouraging results, the development and deployment of robust, safe, and functionally viable human–robot interactive systems remain a major challenge, particularly due to the complexity and unpredictability of outdoor agricultural settings. This thesis addresses one of the central challenges in the field: the integration of robotic platforms into real-world agricultural operations. A significant barrier to widespread adoption lies in the technical expertise required for operating and maintaining such systems, especially in areas where many agricultural workers may lack sufficient training. To overcome this gap, the research proposes a methodology centered on the incorporation of machine learning algorithms to facilitate intuitive and accessible human–robot interaction. The system design emphasizes user-centered interfaces, employing natural human gestures and body movements as a means to control robotic behavior, thereby lowering the barrier for non-expert users and enhancing operational usability in the field. The first implementation presented in this thesis introduces a foundational gesture-recognition system to enable natural and intuitive communication between human operators and robotic platforms. This approach is particularly suited to the unpredictable and constantly changing conditions typical of agricultural environments. The implementation pursues two main objectives: (i) the development of a real-time, skeleton-based hand gesture recognition system capable of identifying five predefined gestures using a combination of machine learning and computer vision techniques; and (ii) the integration and evaluation of this recognition system within a functional HRI framework deployed in diverse real-world agricultural scenarios. Utilizing advanced platforms such as the Robot Operating System (ROS), the system translates recognized gestures into robot-executable commands. Field experiments, conducted in outdoor settings and involving one or two human participants, demonstrated the system’s capability to distinguish between individuals, follow designated users, assist, stop, navigate toward specified targets or disengage based on user intent. Importantly, the robot maintained safe operation throughout all interactions, adhering to socially accepted speeds and preserving appropriate distances from human collaborators. The second phase of the research builds upon this foundation, incorporating recent advancements in robotics, sensor technology, and deep learning to further extend the capabilities of HRI in agriculture. Guided by the evolving field of Agriculture 4.0, this implementation seeks to modernize traditional practices through the integration of real-time data acquisition and adaptive robotic behavior. The objectives of this phase are threefold: (i) to enhance the gesture recognition framework by expanding its gesture repertoire to include more complex and natural interactions; (ii) to establish a more fluid and seamless collaborative relationship between human workers and autonomous robotic systems; and (iii) to facilitate routine agricultural operations through intuitive, context-aware robot behavior. The proposed system was validated through a series of pilot experiments conducted in orchard environments, with the involvement of multiple human participants. These experiments demonstrated the system’s ability to adapt to multi-user scenarios, execute tasks reliably, and support day-to-day agricultural functions, thereby underscoring its practical viability and future potential within the domain of intelligent agriculture.
Degree
thesis:*- Grantor dc:publisher
- University of Thessaly (UTH)
- Year dc:date
- 2025
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Moysiadis, Vasileios
- Μωυσιάδης, Βασίλειος
Subjects
dc:subject × 18- Αλληλεπίδραση ανθρώπου - ρομπότ
- Γεωργία ακρίβειας
- Μηχανική μάθηση
- Νευρωνικά δίκτυα LSTM
- Μηχανική όραση
- Αυτόνομη πλοήγηση
- Ρομποτική
- Human–robot interaction
- Precision agriculture
- Machine learning
- LSTM
- Machine vision
- Autonomous navigation
- Robotics
- Επιστήμη Ηλεκτρολόγου Μηχανικού, Ηλεκτρονικού Μηχανικού, Μηχανικού Η/Υ
- Επιστήμες Μηχανικού και Τεχνολογία
- Electrical Engineering, Electronic Engineering, Information Engineering
- Engineering and Technology
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/59586