来源:https://thinklab-sjtu.github.io/CornerCaseRepo/ , Think2Drive: Efficient Reinforcement Learning by Thinking with Latent World Model for Autonomous Driving (in CARLA-v2)
来源:https://arxiv.org/pdf/2309.05527, ReSimAD: Zero-Shot 3D Domain Transfer for Autonomous Driving with Source Reconstruction and Target Simulation
论文:
Q. Li, X. Jia, S. Wang,
Junchi Yan
(correspondence). Think2Drive: Efficient Reinforcement Learning by Thinking in Latent World Model for Quasi-Realistic Autonomous Driving (in CARLA-v2).
European Conference on Computer Vision (
ECCV
), 2024. https://arxiv.org/pdf/2402.16720
Z. Zhao, F. Fan, W. Liao,
Junchi Yan
.
Grounding and Enhancing Grid-based Models for Neural Fields.
IEEE Conference on Computer Vision and Pattern Recognition (
CVPR
), 2024 Best Paper Candidate
Y. Li, J. Guo (本科生), R. Wang, H. Zha,
Junchi Yan
(correspondence)
OptCM: The Optimization Consistency Models for Solving Combinatorial Problems in Few Shots ,
Neural Information Processing Systems (
NeurIPS
), 2024
斯坦福自然语言处理(NLP)组:成员包括 Chris Manning、Dan Jurafsky、Percy Liang 等。
斯坦福视觉与学习实验室(SVL):由李飞飞、Juan Carlos Niebles、Silvio Savarese、Jiajun Wu 组成。
斯坦福统计机器学习(statsml)组:有 Emma Brunskill、John Duchi、Stefano Ermon 等成员。
研究成果:
来源:https://arxiv.org/pdf/2401.18059, RAPTOR: Recursive Abstractive Processing for Tree-Organized Retrieval
来源:https://arxiv.org/pdf/2404.13026 , PhysDreamer: Physics-Based Interaction with 3D Objects via Video Generation
论文:
TRANSIC: Sim-to-Real Policy Transfer by Learning from Online Correction,
CoRL 2024,
Yunfan Jiang, Chen Wang, Ruohan Zhang, Jiajun Wu, Li Fei-Fei, https://arxiv.org/abs/2405.10315
D3Fields: Dynamic 3D Descriptor Fields for Zero-Shot Generalizable Rearrangement,
CoRL 2024 (Oral),
Yixuan Wang*, Mingtong Zhang*, Zhuoran Li*, Tarik Kelestemur, Katherine Rose Driggs-Campbell, Jiajun Wu, Li Fei-Fei, Yunzhu Li, https://robopil.github.io/d3fields/d3fields.pdf
PhysDreamer: Physics-Based Interaction with 3D Objects via Video Generation,
ECCV 2024 (Oral),
Tianyuan Zhang, Hong-Xing Yu, Rundi Wu, Brandon Y. Feng, Changxi Zheng, Noah Snavely, Jiajun Wu, William T. Freeman, https://arxiv.org/abs/2404.13026
Towards Practical Multi-object Manipulation using Relational Reinforcement Learning, https://richardrl.github.io/relational-rl/
Adversarially trained neural representations may already be as robust as corresponding biological neural representations, https://proceedings.mlr.press/v162/guo22d/guo22d.pdf
Diagnosis, Feedback, Adaptation: A Human-in-the-Loop Framework for Test-Time Policy Adaptation, https://arxiv.org/pdf/2307.06333
Ishida, M., Berio, F., Di Santo, V., Shubin, NH., Iida, F. (2024). Paleo-inspired robotics as an experimental approach to the history of life,
Science Robotics
, (accepted).
Fan, C., Chu, KF., Wang, X., Kwok, KW., Iida, F. (2024). State transition learning with limited data for safe control of switched nonlinear systems,
Neural Networks
, (accepted).
Xu, J., Anvo, NZR., Taha-Abdalgadir, H., d’Avigneau, AL., Palin, D., Wei, F., Hadjidemetriou, F., Iida, F., Al-Tabbaa, A., De Silva, L., Brilakis, I. (2024). Highway digital twin-enabled Autonomous Maintenance Plant (AMP): A perspective,
Data-Centric Engineering
, (accepted).
Almanzor, E., Sugiyama, T., Abdulali, A., Hayashibe, M., Iida, F., (2024). Utilising redundancy in musculoskeletal systems for adaptivestiffness and muscle failure compensation: A model-free inverse statics approach,
Bioinspiration & Biomimetics
19: 046015.
Affective Intelligence and Robotics Laboratory (AFAR)-剑桥
Learning Socially Appropriate Robo-waiter Behaviours through Real-time User Feedback, https://dl.acm.org/doi/10.5555/3523760.3523831
Domain-Incremental Continual Learning for Mitigating Bias in Facial Expression and Action Unit Recognition, https://ieeexplore.ieee.org/document/9792455
Latent Generative Replay for Resource-Efficient Continual Learning of Facial Expressions, https://www.repository.cam.ac.uk/items/ca5b5996-350c-4354-9f5c-941bcc16224b
Oxford Robotics Institute
主页:https://ori.ox.ac.uk/
ORI 致力于拓展机器人的能力,为人类带来更多价值。该研究所的研究兴趣极为广泛,涵盖了从飞行到抓取、从检查到奔跑、从触觉到驾驶、从探索到规划等众多领域。在技术研究方面,ORI 涉及机器学习与人工智能、计算机视觉、制造、多光谱传感、感知、系统工程等多个前沿主题。
研究成果:
论文:
Textual explanations for automated commentary driving, https://ieeexplore.ieee.org/document/10186611
Motion planning in dynamic environments using context-aware human trajectory prediction, https://www.sciencedirect.com/science/article/pii/S0921889023000891?via%3Dihub
EDAMS: An Encoder-Decoder Architecture for Multi-grasp Soft Sensing Object Recognition, https://ieeexplore.ieee.org/document/10121962
Marter, P., Khramova, M., Duvigneau, F., Wood, R.J., Juhre, D. and Orszulik, R., 2024. Bidirectional motion of a planar fabricated piezoelectric motor based on unimorph arms. Sensors and Actuators A: Physical, 377, p.115642. October 2024, https://doi.org/10.1016/j.sna.2024.115642
Burns, J.A., Daniels, J., Becker, K.P., Casagrande, D., Roberts, P., Orenstein, E., Vogt, D.M., Teoh, Z.E., Wood, R., Yin, A.H. and Genot, B., 2024. Transcriptome sequencing of seven deep marine invertebrates. Scientific Data, 11(1), p.679. June 2024, https://doi.org/10.1038/s41597-024-03533-4
Maalouf, A., Jadhav, N., Jatavallabhula, K.M., Chahine, M., Vogt, D.M., Wood, R.J., Torralba, A. and Rus, D., 2024. Follow Anything: Open-set detection, tracking, and following in real-time. IEEE Robotics and Automation Letters, 9(4), pp.3283-3290, April 2024, doi: 10.1109/LRA.2024.3366013.
(3)生成式 AI:构建生成式 AI 模型,使模型能够用自然语言解释医学图像,并实现与临床医生的交互通信。引入数据集以解决自动报告生成的一些重大挑战,还开创了放射学报告生成的辅助方法,研究为临床医生提供 AI 辅助的影响以及可解释性方法的可信度。
论文:
Expert-level detection of pathologies from unannotated chest X-ray images via self-supervised learning, https://www.nature.com/articles/s41551-022-00936-9
MoCo-CXR: MoCo Pretraining Improves Representation and Transferability of Chest X-ray Models, https://arxiv.org/abs/2010.05352
Predicting patient decompensation from continuous physiologic monitoring in the emergency department, https://www.nature.com/articles/s41746-023-00803-0
Robotics and Embodied Artificial Intelligence Lab (REAL)
主页:https://real.stanford.edu/index.html
研究方向:多机器人协作与语言模型、机器人技能学习与获取、机器人操作研究、机器人导航与场景理解
研究成果:
来源:https://maniwav.github.io/, ManiWAV: Learning Robot Manipulation from In-the-Wild Audio-Visual Data
来源:https://umi-on-legs.github.io/, https://www.youtube.com/watch?v=4Bp0q3xHTxE. UMI 是一个带有 GoPro 相机的手持式夹具。有了 UMI,我们可以在任何地方为任何操作技能收集现实世界中的机器人演示数据,而且不需要任何机器人。所以,只要带着 UMI 走到户外,就可以开始收集数据啦!
论文:
GET-Zero: Graph Embodiment Transformer for Zero-shot Embodiment Generalization, https://arxiv.org/abs/2407.15002, https://get-zero-paper.github.io/
Dynamics-Guided Diffusion Model for Robot Manipulator Design, https://dgdm-robot.github.io/ , https://arxiv.org/abs/2402.15038
Multi - Scale Embodied Intelligence Lab 由 Dandan Zhang 博士领导,专注于多尺度的具身智能研究,将机器学习与机器人技术相结合,主要针对医疗应用领域,目标是开发具有超人类能力的下一代机器人。
研究方向:(1)多模态感知(传感):研究如何让机器人在不同尺度下精确地感知环境,例如针对微尺度手术工具的感知研究。(2)交互学习(决策):通过机器学习等技术,探索机器人如何在复杂环境中进行决策,如结合人工操作指令和基于机器学习的自主控制实现高效的人机共享控制。(3)灵巧操作(执行):研究机器人在不同尺度下的灵巧操作能力,包括开发微手术工具以实现精确的微操作,以及研究在 3D 空间中对微工具的灵巧操作控制策略和视觉技术。
应用领域:包括医疗机器人、家用机器人和辅助机器人等。
研究成果:
来源:https://arxiv.org/pdf/2303.02708 , Tac-VGNN: A Voronoi Graph Neural Network for Pose-Based Tactile Servoing
来源:https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10155191 , TacMMs: Tactile Mobile Manipulators for Warehouse Automation
来源:https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10182274 , Sim-to-Real Model-Based and Model-Free Deep Reinforcement Learning for Tactile Pushing
论文:
D. Zhang
*, J. Zheng, "
Towards the New Generation of Smart Homecare with IoHIRT: Internet of Humans and Intelligent Robotic Things
",
under review
.
D. Zhang
*, Z. Wu, J. Zheng, Y. Li, Z. Dong, J. Lin, "
HuBotVerse: A Mixed Reality-Aided Cloud-Based Framework
",
under revision
.
W. Fan*, H. Li*, W. Si, S. Luo, N. Lepora,
D. Zhang
,
* "
ViTacTip: Design and Verification of a Novel Biomimetic Physical Vision-Tactile Fusion Sensor
",
under review
.
Sensing, Interaction & Perception Lab - SIPLAB - ETH Zürich
EgoPoser: Robust Real-Time Egocentric Pose Estimation from Sparse and Intermittent Observations Everywhere. Jiaxi Jiang, Paul Streli, Manuel Meier, and Christian Holz.European Conference on Computer Vision 2024 (ECCV).
Ultra Inertial Poser: Scalable Motion Capture and Tracking from Sparse Inertial Sensors and Ultra-Wideband Ranging. Rayan Armani, Changlin Qian, Jiaxi Jiang, and Christian Holz.Proceedings of ACM SIGGRAPH 2024.
Robust Heart Rate Detection via Multi-Site Photoplethysmography. Manuel Meier and Christian Holz.Proceedings of IEEE EMBC 2024.
Robot Perception and Learning Lab
主页:https://rpl-as-ucl.github.io/
Robot Perception and Learning Lab 是一个专注于机器人研究的实验室,工作重点是为关节式机器人(四足、人形、类动物、移动操作器等)开发前沿算法,使它们能够在不可预测的自然环境中高效导航和操作。通过结合感知、决策和运动控制方面的见解,旨在使机器人能够以更高的自主性和精确性处理现实世界场景的复杂性。