School of Engineering Department of Electronic and Computer Engineering 172 Development of Bioinspired Tactile Sensor Supervisor: SHEN Yajing / ECE Student: LIU Yixuan / ELEC Course: UROP 1100, Fall This UROP focuses on completing a neural network which is able to estimate the roughness of some fabric textures. The neural network, based on Pytorch platform, is a kind of classic CNN model. The key aspects include the model construction, data acquisition, parameters adjustment and model training. The goal is to create a neural network with the input of friction data and the output of parameters of texture. My research this time is aimed at coding a simplified basic model and collecting the different friction data. In this way, the model can help estimate the level of roughness according to the friction data input. Development of Bioinspired Tactile Sensor Supervisor: SHEN Yajing / ECE Student: TONG Zhe / CPEG Course: UROP 1100, Fall In this Undergraduate Research Opportunities Program (UROP) project, I concentrated my efforts on the innovative development of a tactile sensor that incorporates both self-decoupling and super-resolution capabilities. This endeavor was rooted in a previous design of a soft-based sensor, which provided a foundational understanding of the principles involved in tactile sensing. Building on that foundation, I embarked on the creation of a series of advanced sensors, alongside a comprehensive testing platform that would allow for the precise application of forces at predetermined positions. Development of Bioinspired Tactile Sensor Supervisor: SHEN Yajing / ECE Student: YEUNG Wun Lam / CPEG Course: UROP 1100, Spring Robotic hand exoskeletons restore hand function for individuals with motor impairments and enhance performance in tasks like typing or playing instruments. However, most systems focus on basic flexion and extension, lacking adaptive grasping for various object sizes and control of finger abduction and adduction, critical for complex activities. This research introduces two systems: the first provides six degrees of freedom with vision-based object size estimation, adaptive grasp type selection, real-time web control, and PID-based pressure sensing. The second adds independent abduction and adduction control, increasing the total degrees of freedom to eight, enabling advanced tasks like typing. These advancements mark significant progress in multifunctional, user-adaptive hand exoskeletons for daily assistance and enhanced performance.
RkJQdWJsaXNoZXIy NDk5Njg=