2026 ASEE Annual Conference & Exposition

Traditional exoskeleton technologies have focused primarily on enhancing existing human capabilities or restoring lost key motor functions. While these systems successfully assist movement, they are rarely designed to help users maintain or gradually improve muscle strength, or to align their movement habits over time. This project addresses these limitations through artificial intelligence (AI)-based decision-making that enables adaptive, user-centered control. Building on an exoskeleton arm previously presented at the 2025 ASEE Annual Conference, the project integrates mechanical, electrical, and computational components to form a complete mechatronics system enhanced by modern machine-learning techniques. The exoskeleton consists of two pipe-like braces attached to the forearm and upper arm, connected by a motorized joint driven by a microcontroller-based control system. A load cell mounted on the frame measures the tension produced by the bicep muscle and serves as the primary input for control and classification. Using load cell data, a machine-learning classifier trained through Edge Impulse determines user intent in real time. The classifier was deployed as a lightweight edge-AI model on a Raspberry Pi and achieved an accuracy of 75.7% in distinguishing action patterns such as picking and holding versus picking and dropping. To improve motion quality, a DC motor and closed-loop proportional-integral (PI) control algorithm were implemented, resulting in smoother and more stable arm movement compared to the previous servo-based design. Conducted by a high school student, this project encompasses all aspects of engineering design, hardware development, algorithm training, coding, and testing. It demonstrates how pre-college students can meaningfully engage in authentic STEM research that connects mechatronics, AI, and human-centered design.