Embodied AI for Complex and Dynamic Scenarios

Abstract: Dynamic interactive scenarios in complex real-world settings place extremely high demands on robots’ environmental adaptability, task generalization and real-time response capabilities. Focusing on three core modules—embodied perception, cognitive reasoning and behavioral execution—this talk explores key technical frameworks and implementation solutions for complex dynamic interaction scenarios. Leveraging end-to-end large model techniques, it further analyzes the paradigm transformation and future development trends of embodied intelligence.

Yuexin Ma is Tenured Associate Professor at ShanghaiTech University. She received her Ph.D. from the University of Hong Kong. Her research focuses on 3D vision, embodied intelligence, and autonomous driving. She has authored over 100 papers in top-tier journals and conferences of relevant fields, including more than 50 first-author and corresponding-author publications such as TPAMI, CVPR, ICCV, ICML, AAAI, etc. Her Google Scholar citations exceed 9,000. Papers supervised by her won the Best Paper Award (Single Recipient) at MICCAI 2024 and were selected as Best Paper Candidates at ACM MM 2024. Her honors include Shanghai Overseas High-Level Talent, First-Class Prize of Shanghai Excellent Teaching Achievement, Outstanding Young Researcher Award at China 3DV 2025, and inclusion in the World’s Top 2% Scientists List.