How AI and Automation Are Reshaping Shipyard Labor Structures 6 views

The integration of artificial intelligence into physical labor environments forces us to reevaluate how human workers interact with an increasingly automated world. My technical capstone project details the development and training of a prototype autonomous shipyard robot, serving as a proof of concept for utilizing vision-language-action models to automate heavy industrial tasks. Correspondingly, my Science, Technology, and Society research examines the sociotechnical implications of this shipyard automation to understand how human workers adapt to these systems. These two projects are fundamentally connected by their shared focus on the realities of implementing artificial intelligence in industrial settings. While the technical project addresses the physical and computational challenges of building an autonomous system, the STS paper critically analyzes the human and societal impacts of integrating such technology onto the factory floor. The capstone project was designed to address the need for improved safety and efficiency for industrial military contractors building ships for the United States Navy. By creating a semi-autonomous system for hazardous work, the project contributes to reducing workplace accidents and human error. The methodology involved modifying a Hugging Face LeKiwi robotic arm with custom printed circuit boards to control a mounted green laser diode and an active warning buzzer. A lightweight smolVLA machine learning model was then trained using teleoperation to teach the system how to autonomously identify, grasp, and relocate a target object within a scaled-down shipyard environment. Ultimately, the capstone project demonstrated that the modified robot could successfully perform complex industrial tasks. After iterative dataset refinement, a 180-episode smolVLA model achieved a 75% accuracy rate in acquiring and relocating the target object, and the custom safety modifications met all functional specifications. However, the research also revealed the limitations of current vision-based automation, highlighting that these models remain heavily dependent on vast amounts of formatted training data and static environmental conditions to operate effectively. For the STS paper, the primary research question explores how autonomous robotic agents reshape the sociotechnical landscape of the industrial shipyard. This is highly significant because shipyard workers must navigate the integration of these machines in a way that enhances their roles rather than threatening their safety or livelihoods. The methodology applies the Social Construction of Technology and Actor-Network Theory frameworks to evaluate the negotiations and power dynamics between human laborers, corporate stakeholders, and the artificial intelligence models themselves as non-human actors. The evidence draws upon the historical integration of semi-autonomous systems in labor-intensive industries, contrasting generative artificial intelligence's impact on white-collar work with its physical manifestation in blue-collar environments. The results indicate that the success of robotic automation relies just as heavily on open communication and worker acceptance as it does on accurate machine learning datasets. I conclude that for shipyard automation to be truly effective, the engineering design process must continuously account for the human network it disrupts, ensuring that technological progress is socially constructed to protect and assist the worker.

BS (Bachelor of Science)

AI; Shipyard; Automation

English

2026-05-07