School of Engineering Department of Electronic and Computer Engineering 170 Embodied Robotic Arm Systems with AI-Based Control Policies Supervisor: SHAO Qiming / ECE Student: HUANG Yuen Ngai / COMP Course: UROP 1000, Summer The field of embodied artificial intelligence is a rapidly growing area of research and development, with a focus on the field of robot control. This paper provides a literature review on two AI control policies based on the diffusion model and evaluates their performance in simulation and real-world scenarios. The paper also provides a review of related literature and present the results of replication as an effort to dive deeper into the field. Embodied Robotic Arm Systems with AI-Based Control Policies Supervisor: SHAO Qiming / ECE Student: TONG Zhe / CPEG Course: UROP 1100, Spring This literature review examines AI-based control policies for embodied robotic arm systems, focusing on six representative works employing reinforcement learning (RL), imitation learning (IL), supervised learning (SL), diffusion transformers, and vision-language-action (VLA) models. These studies address tasks like pick-andplace, bimanual manipulation, and inverse kinematics, evaluated in simulated and real-world settings. Key advancements include the Robotics Diffusion Transformer (RDT) for dexterous bimanual tasks, OpenVLA for generalist manipulation with language grounding, and Mobile ALOHA for cost-effective bimanual platforms. The review compares methodologies, applications, and performance metrics, highlighting trade-offs in adaptability, data efficiency, and computational demands. Future directions include hybrid models, improved generalization, and standardized benchmarks for robotic control. AI Enhanced Smart Design, Manufacturing and Robot Supervisor: SHEN Yajing / ECE Student: HUANG Zhenghao / ELEC Course: UROP 2100, Summer Numerous industries, such as the clothing manufacturing industry, the medical field (gauze, etc.), the recycling of textiles, and others, depend on fabric. However, most of these pieces are difficult for robots to handle since they require a strong sense of touch. Once robots can substitute human people for such jobs, it can bring about several advantages, such us raising productivity and cutting labor expenses. The goal of this research is to develop a revolutionary fabric pinching system (FPS) that can identify and separate a single fabric layer. The gripper’s design was influenced by how people rub their hands. Thus, I attempted to simulate the rubbing motion of human hands by using a rolling method. The details will be elaborated in the following text.
RkJQdWJsaXNoZXIy NDk5Njg=