publications

2026

1. ECCV

   AutoSpeed: Annotation-Free Stage-Adaptive Motion Speed Learning for Robot Manipulation

   Qingda Hu^†, Ziheng Qiu^†, Jieru Zhao, Zhongxue Gan, and Wenchao Ding

   In European Conference on Computer Vision, 2026

   Abs arXiv Bib Code Website

   AutoSpeed is an annotation-free, model-agnostic framework that learns stage-adaptive motion speeds for visuomotor policies. It uses cost-aware selection over DCT-retimed trajectory targets and Nonlinear Temporal Aggregation for smooth deployment. Across 62 simulation tasks and four real-world bimanual tasks, AutoSpeed reduces execution time while maintaining or improving task success.

   @inproceedings{hu2026autospeed,
     title = {AutoSpeed: Annotation-Free Stage-Adaptive Motion Speed Learning for Robot Manipulation},
     author = {Hu, Qingda and Qiu, {<strong>Ziheng</strong>} and Zhao, Jieru and Gan, Zhongxue and Ding, Wenchao},
     booktitle = {European Conference on Computer Vision},
     year = {2026},
   }

2. RA-L

   Resolving State Ambiguity in Robot Manipulation via Adaptive Working Memory Recoding

   Qingda Hu, Ziheng Qiu, Zijun Xu, Kaizhao Zhang, Xizhou Bu, Zuolei Sun, Bo Zhang, Jieru Zhao, Zhongxue Gan, and Wenchao Ding

   IEEE Robotics and Automation Letters, 2026

   Abs arXiv Bib Code Website

   PAM is a visuomotor policy with adaptive working memory for resolving state ambiguity in robot manipulation. It recodes multi-scale history into compact context features, supports a 300-frame history window, and runs above 20 Hz. A single multi-task policy reaches a 0.91 average success rate across seven real-world tasks.

   @article{hu2026pam,
     title = {Resolving State Ambiguity in Robot Manipulation via Adaptive Working Memory Recoding},
     author = {Hu, Qingda and Qiu, {<strong>Ziheng</strong>} and Xu, Zijun and Zhang, Kaizhao and Bu, Xizhou and Sun, Zuolei and Zhang, Bo and Zhao, Jieru and Gan, Zhongxue and Ding, Wenchao},
     journal = {IEEE Robotics and Automation Letters},
     year = {2026},
   }