Zexian Ji is an MPhil student in the Department of Mechanical Engineering at The Hong Kong Polytechnic University. He earned his bachelor's degree from Harbin Engineering University. His research focuses on robotics, with a particular emphasis on imitation learning for robotic systems. Zexian’s work explores the application of generative models, such as diffusion models, in trajectory planning for robotic manipulation. The ability of diffusion models to capture multi-modal distributions is especially advantageous in tasks where multiple equally valid solutions (redundant solutions) exist, such as trajectory planning and grasping. By capturing these diverse solutions, his research aims to enhance the generalizability and versatility of robots, enabling them to generate feasible solutions under varying conditions, such as different object placements or constraints during inference.