Vision-Based Human Intention Recognition and Robotic Assistance for Smart Work Cells
Year: 2025
Type: dissertation
Abstract: Human-robot collaboration (HRC) plays a vital role in enhancing productivity and inclusivity in modern work environments. However, existing systems often fail to address the specific needs of individuals with physical disabilities. This dissertation proposes an intelligent, vision-based robotic assistance system tailored for inclusive workcells, specifically supporting users with only one functional arm. The system integrates a depth camera, a gimbal-based motion tracking mechanism, and MediaPipe-based human pose and hand tracking, combined with lightweight machine learning models for real-time human intention recognition. The robotic platform centers around a UR3e collaborative arm equipped with modular, 3D-printed end-effectors designed for flexible object handling and soldering tasks. The system supports both manual and autonomous operation modes, enabling seamless gesture-to-action mapping without requiring wearable devices. Real-time gesture classification is achieved with high accuracy using ensemble learning techniques, demonstrating over 90% recognition rates in dynamic settings. Comprehensive experiments validate the system’s robustness across diverse user conditions, including gesture variability, spatial movement, and real-time task execution. A gimbal-stabilized vision pipeline ensures consistent tracking, and safety features are integrated to prevent unintended robot actions. While initial tests were conducted with healthy users, the system is designed for older adults and persons with disabilities, and future studies will expand its evaluation in those populations. This work bridges theoretical advancements in vision-based intention recognition and practical deployment of assistive robotics. It contributes a scalable and adaptable framework for inclusive automation, addressing critical challenges in aging societies and paving the way for future human-centered manufacturing.
DOI: 10.18997/0002002145
Author: Suppa-Adirek, Natchanon
Language: English
Subfield: Human-Computer Interaction
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7116044498
Cites:
Cited by:
Related to: