Gait Event Detection and Torque Adjustment for Multi-Terrain Locomotion

Affiliations:	College of Engineering and Computer Science
Team Leader:	Sy Nguyen sy.nguyen@ucf.edu Mechanical & Aerospace Engineering
Faculty Mentor:	Hwan Choi, PhD
Team Size:	10
Open Spots:	5
Team Member Qualifications:	We are seeking motivated undergraduate team members who are interested in biomechanics, prosthetics, and machine learning. The following skills or experiences are preferred but not required: + Major in Biomedical Engineering, Mechanical Engineering, or related field +Experience with musculoskeletal modeling software, especially OpenSim +Familiarity with deep learning frameworks (e.g., TensorFlow, PyTorch) +Basic understanding of biomechanics, gait analysis, or human movement science +Interest in applying AI to human motion and rehabilitation technologies.
Description:	This project aims to develop a robust gait detection and torque adjustment system capable of adapting to diverse walking terrains. The system integrates inertial measurement unit (IMU) sensors and vision-based camera data to accurately identify gait events, gait phases, and terrain transitions during human locomotion. IMU sensors mounted on the lower limb provide high-frequency motion data, including linear acceleration and angular velocity, which are used to detect key gait events such as heel strike, toe-off, and stance–swing transitions. Complementary camera-based perception is employed to recognize terrain types and environmental context, enabling early detection of upcoming terrain changes. The considered terrains include concrete, grass, sand, ascending stairs, descending stairs, and sloped surfaces. By fusing IMU and vision data, the proposed system enhances robustness and generalization across varying walking conditions. Detected gait events and terrain information are used to dynamically adjust joint torque profiles, allowing the system to provide terrain-appropriate assistance or control. This adaptive torque adjustment aims to improve walking stability, efficiency, and safety, particularly in unstructured environments. The developed framework is intended for application in lower-limb prostheses, wearable robotic systems, or assistive exoskeletons, where reliable gait detection and terrain-adaptive control are critical. The results of this project are expected to contribute to more intelligent and responsive locomotion systems capable of seamless multi-terrain walking.