4D Scene Understanding for Intelligent Robotics

About the Workshop

Modern robotic systems must operate safely and robustly in dynamic environments. Conventional methods for tracking, mapping, grasping, view synthesis, and scene understanding often struggle when assumptions about scene stability are violated. This workshop explores the challenges and solutions of 4D robotics—spatial perception enhanced by temporal context—by exploring the interplay of dynamic localization, real-time reconstruction, high-fidelity rendering, and generative modeling. We aim to identify and standardize benchmarks, discuss novel evaluation protocols, and inspire cross-disciplinary collaborations. Through engaging keynotes and paper presentations, participants will gain insights into state-of-the-art approaches and chart future research directions for next-generation robotics.

Organizers

Yanyan Li

National University of Singapore

Xin Li

Nanyang Technological University

Yan Di

Technology University of Munich
Harbin Institute of Technology

Ruida Zhang

Tsinghua University

Xiangyue Liu

Hong Kong University of Science and Technology

Yida Wang

Li Auto

Zhongliang Jiang

Technology University of Munich

Chau Yuen

Nanyang Technological University

Contact: yanyan.li.camp@gmail.com; xin019@e.ntu.edu.sg

Invited Speakers

Speaker

TBD

We are also engaging additional speakers from diverse geographical regions and underrepresented groups to ensure a broad set of perspectives.

Tentative Schedule

Call for Papers

Submissions are non-archival, allowing previously accepted or ongoing work to be presented (accepted papers will be made available on our workshop website). Three categories of papers are welcomed:

• Full Papers: Up to 8 pages (excluding references and appendix, i.e., 8+n pages)
• Short Papers: Up to 4 pages (excluding references and appendix, i.e., 4+n pages)
• Extended Abstracts: Up to 2 pages (excluding references and appendix, i.e., 2+n pages)

We will conduct a double-blind review with at least two expert reviewers per paper. Submissions will be judged on relevance, technical merit, clarity, and potential impact. Outstanding papers will be invited for extended submissions to a special journal issue (TBD). Two Best Paper Awards (oral and poster) are planned to recognize outstanding papers and encourage high-quality contributions.

Important Dates(Tentative)