Sixu Yan | 鄢思旭

Email  /  Google Scholar  /  Github /  Twitter /  WeChat

@article{zhang2025m3bench,
title = {M${}^{3}$Bench: Benchmarking Whole-Body Motion Generation for Mobile Manipulation in 3D Scenes},
author = {Zhang, Zeyu and Yan, Sixu and Han, Muzhi and Wang, Zaijin and Wang, Xinggang and Zhu, Song-Chun and Liu, Hangxin},
journal = {IEEE Robotics and Automation Letters},
year = {2025},
volume = {10},
number = {7},
pages = {7286-7293},
publisher = {IEEE}
}

We propose M³Bench, a large-scale benchmark and data generation toolkit for evaluating whole-body motion generation in mobile manipulation. It includes over 30,000 pick-and-place tasks across 119 realistic 3D scenes, with expert trajectories generated by VKC planner. Our benchmark supports assessing generalization to novel scenes and objects, and reveals that existing methods struggle with whole-body coordination under task and environmental constraints.

@article{yan2025m2diffuser,
title = {M2Diffuser: Diffusion-based Trajectory Optimization for Mobile Manipulation in 3D Scenes},
author = {Yan, Sixu and Zhang, Zeyu and Han, Muzhi and Wang, Zaijin and Xie, Qi and Li, Zhitian and Li, Zhehan and Liu, Hangxin and Wang, Xinggang and Zhu, Song-Chun},
journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence},
year = {2025},
publisher = {IEEE}
}

We propose M²Diffuser (Mobile Manipulation Diffuser), a conditional diffusion-based neural motion planner capable of generating full-body coordinated trajectories that satisfy both physical and task constraints. This approach provides a unified framework for motion generation and trajectory optimization in 3D scenes and enables seamless transfer from simulation to real-world robotic platforms.

@InProceedings{liao2025diffusiondrive,
author = {Liao, Bencheng and Chen, Shaoyu and Yin, Haoran and Jiang, Bo and Wang, Cheng and Yan, Sixu and Zhang, Xinbang and Li, Xiangyu and Zhang, Ying and Zhang, Qian and Wang, Xinggang},
title = {DiffusionDrive: Truncated Diffusion Model for End-to-End Autonomous Driving},
booktitle = {Proceedings of the Computer Vision and Pattern Recognition Conference (CVPR)},
year = {2025},
pages = {12037-12047}
}

We propose DiffusionDrive, a truncated diffusion-based planner for real-time end-to-end autonomous driving. By injecting multi-mode anchors and reducing the denoising steps to two, it achieves fast inference while maintaining high-quality trajectory prediction. DiffusionDrive outperforms prior methods on the NAVSIM benchmark in both planning accuracy and diversity.