一、主讲人: 黄廷文 深圳理工大学讲席教授、长江讲座教授

二、讲座题目：AI-Driven Motion Imitation in Dinosaur Models via Neural and Modal Analysis

三、地   点：10110

四、时   间：2025年12月14日（周日）10:00-11:00

五、主办部门：ok138太阳集团古天乐

欢迎广大师生参与本次讲座！请参加讲座人员于2025年12月14日9:50前入场。

报告摘要：In this talk, we present an AI-based framework for motion imitation between distinct 3D dinosaur models using modal analysis and neural networks. By decomposing complex animations into modal coordinates via eigenmode analysis of finite element models, we obtain interpretable and compact motion representations. Specifically, we extract modal coefficients over time from a reference dinosaur (e.g., a T-Rex) and aim to transfer the motion to a target dinosaur (e.g., a Brachiosaurus) with different morphology. To learn this cross-modal mapping, we construct a synthetic dataset of modal coefficient sequences and employ Long Short-Term Memory (LSTM) networks to predict the target dinosaur’s modal dynamics from those of the source. We further explore the integration of Physics-Informed Neural Networks (PINNs) to enforce structural consistency during inference. Our experiments demonstrate that the LSTM can successfully learn to approximate complex motion behaviors across morphologically distinct models, and the addition of physics constraints further improves stability and realism. This work bridges physical modeling and AI, offering a promising methodology for animating biomechanical systems with minimal manual rigging or retargeting.

个人简介：黄廷文是深圳理工大学讲席教授。他于1990年毕业于西南师范大学,现西南大学，获学士学位；1993年毕业于四川大学，获硕士学位；2002年获得德州农工大学博士学位。他于2003年加入德州农工大学卡塔尔分校担任助理教授，2013年晋升为教授。他的研究领域包括神经网络、动力系统、复杂网络、优化与控制、智能电网, 多智能体。他是长江讲座教授，亚太人工智能协会(AAIA) Fellow, IEEE Fellow, 国际模式识别协会(IAPR) Fellow, 国际系统与控制科学院院士, 欧洲科学与艺术院院士, 发展中国家科学院院士。