EEG-Controlled Robotics; Barriers to Medical Device Distribution and Effectiveness in the U.S.A. 13 views

For many patients, state-of-the-art prosthetic and assistive devices are practically inaccessible. To develop a system that uses electroencephalography (EEG) signals and reinforcement learning to control a robotic arm in real time, the research team investigated brain-computer interfacing (BCI) as a more accessible form of prosthetic control for patients who cannot rely on muscle-based inputs. The final system used an OpenBCI EEG headset, a Raspberry Pi, and a 3D-printed robotic arm with one degree of freedom. A Deep Q-Network reinforcement learning algorithm was trained to distinguish between tensing and relaxing brainwave patterns and translate them into motion. In 100 trials, the system recognized user intent correctly in about 88% of cases, showing that reinforcement learning can improve responsiveness and reliability compared to more rigid machine-learning methods. Administrative and policy constraints limit access to advanced prosthetic and medical devices in the United States. The Social Construction of Technology (SCOT) framework reveals how participant groups, including manufacturers, clinicians, insurers, regulators, and advocates, define value differently. These misaligned definitions determine which devices are made available for use. In practice, access is determined not only by the technology itself, but also by how reimbursement standards and health policy value it.

BS (Bachelor of Science)

EEG; Prosthetic; Reinforcement Learning; Medical Device

School of Engineering and Applied Science Bachelor of Science in Mechanical Engineering Technical Advisor: SarahSun STS Advisors: Travis Elliot, Peter Norton Technical Team Members: Cayla Celis, Abigail Dodd, Joshua Rivas-Zelaya,

English

All rights reserved by the author (no additional license for public reuse)

2025-12-12