• 2 Papers were accepted to ICRA 2022
    (March 2022, ICRA)
    Two papers in non-holonimic grasping and dig-grasping were published in cooperation with Prof. Jungwon Seo at HKUST.
  • Visual-Semantic Graph Attention Network for Human-Object Interaction Detecion
    (Oct 2021, ROBIO)
    In this paper, we show how dual visual-semantic GNNs with attention can be leveraged for improved human-object interaction.
  • Our joint NSFC Funding for the General Program was approved
    (Sept 2021)
    The fund in collaboration with Paul Weng at Shanghai Jiatong University, will explore how to make DRL more sample efficient, particularly in physical agents like robots. This would enable to facilitate the training of real robots and increase the adoption of DRL.

  • Towards Safe Control of Continuum ManipulatorUsing Shielded Multiagent Reinforcement Learning
    (July 2021, CASE-RAL)
    In this paper, we show how multi-agent DQN can be used to control a surgical continuum manipulator (simulator and real) with external payloads and disturbances (soft obstacles, rigid collisions) to do point- and trajectory-tracking. A shielding mechanism improved safe control of the robot yielding sub-millimeter RMS errors.

  • University of California Berkeley Collaboration
    (May 2021, HKCLR)
    As a PI in the newly formed Hong Kong Center for Logistic Robotics in HK, Science Park, I am part of a broad and exciting collaboration with UC Berkley to tackle a wide range of problems across the robotics domain!

 

  • GAST (Graph Attention Spatio-Temporal CNNs)
    (May 2021, ICRA)
    In this paper, 3D pose estimation with better interleaving of spatio-temporal dynamics and representation of local and global kinematic constraints in the human skeleton for  2D-to-3D video pose estimation.

 

  • Autotune
    (March 2021, IROS-RAL)
    DRL hyperparameters are difficult to tune. Often leading to wasted time, computational resources, and leading to redundant compute. Check out at how we optimize hyperparameters in self-supervised settings that use VAEs on image space.

     

  • SPAIR (Sense-Plan-Act-Introspect-Recover)
    (Feb 2021, Journal of Intelligent Robot Systems)
    Check out this journal publication on: Identification, Classification, and Recovery from failures with extensive experimental verification. Should be of very high interest to those applying intelligent robots to the unstructured real world. Open-data (including raw data) and open-code.

 

  • Error Recovery and Identification in Robotic Snap Assembly
    (Jan 2021, Symposium of System Integration)
    This joint work with Yusuke and Prof. Harada’s time from Osaka U, got a Best Paper Award Finalist nomination.
  • Neurips DRL Workshop
    (Dec 2020, Neurips)
    An early version of our work on Hyperparameter Auto-tuning in Self-Supervised Robotic Learning was accepted for poster presentation.

  • CUHK
    I started a new position at the Chinese University of Hong Kong.

  • Book Publication
    Hongmin led the publication of his research in a Springer Open book. Take a look.

  • 2020 RAL-IROS
    Our paper titled “Invariant Transform Experience Replay: Data Augmentation for Deep Reinforcement Learning” was accepted for publication in IEEE Robotics and Automation Letters (RAL) and will be presented in IROS 2020.
  • 2019 August RCAR Paper Best Student Paper Finalist & Tutorial
    Our paper on dynamic IProMPs was a finalist for the best student paper award category. We also presented a tutorial on ROS and DRL.
  • 2019 July Do Good Robotics Symposium
    Our paper on computational thinking skills via the ROOT robot was accepted for poster presentation!
  • 2019 Several journal publications on nonparametrics in robotics
    Hongminwu published 3 journals on different studies we conducted on nonparametric Bayesian Markov Switching Processes in robotic anomaly detection and classification.

 

  • 2018 December Humanitarian Technology Paper Accepted
    This one is not in our main area of research, but we thought we could voice the small amount of robot research in the cleanup of oceans given the catastrophic amount of plastic pollution in the oceans.
  • 2018 October ROBIO Paper and Tutorial Accepted
    Our submission “Multimodal Sparse Representation for Anomaly Classification in A Robot Introspection System” was accepted at the 2018 ROBIO conference in Beijing, China. We also presented the tutorial on: “Motion Skill Generation and Classification for Human Robot Interaction”.
  • 2018 July Humanoids Paper Accepted
    Our submission “Collaborative Human-Robot Motion Generation using LSTM-RNN” was accepted at the 2018 Humanoids conference in Beijing, China.
  • 2018 Feb ROMAN TUTORIAL
    Our proposed tutorial for ROMAN 2018 on “Full Day Tutorial on Motion Skill Generation and Classification for Human Robot Interaction” was accepted. Find the page at:
    http://www.juanrojas.net/hri_skill_gen_class_tutorial/
  • 2017 September NSFC Project Approved
    Our project: “Predicting Human Behavior for Efficient Physical Human-Robot Interaction” was approved. The project was funded from January, 2018 to December, 2019 for 320,000RMB under Proj. Number: 61750110521.
  • New ROS MOOC!
    You might be surprised, but this MOOC is hosted by a Chinese MOOC company of the likes of Udacity or Coursera. The course is taught in English, though HW material can be submitted in both English and Mandarin.The courses’ main text is Prof. Wyatt Newman’s newly released text: “A Systematic Approach to Learning Robot Programming with ROS“.  Supplemental texts will be used for reference. We will use Patrick Goebel’s famous ROS By example books, both Vol. 1 (EN/CN) and Vol. 2 (EN/CN). Students worldwide are welcome.The class and text stand out for two reasons: you will be taught by someone who has 14 years of middleware experience, 6 years of ROS class teaching and you will be taught Robotics and ROS  in a systematic way! You will not just be presented with a series of projects stringed together. You will learn principles and use beginning to advanced code examples that build on top each other systematically. From the ROS graph, to simulation, visualization, camera visualization and operations, point clouds, manipulation, and system integration.If you are interested, you can register until Sept. 30. See the online course page and sign up.The slides will also be available independently for purchase at:
    https://www.teacherspayteachers.com/Store/Learn-Ros-And-Robotics-Systematically
  • 2017 June ROS By Example, Vol. 2 Indigo Version Chinese Version Published!
    Patrick’s popular volume has now been officially translated into Chinese. The book is available on all Chinese online stores! Find it in dangdang.com here, amazon.cn here, and jd.com here. This marks our 2nd translation of ROS books into Chinese.
  • 2017 June New Postdoctoral Fellow and Research Associate in the Lab
    We welcome Sakmongkok Chumkamon. He received his Ph.D. from Kyushu Institute of Technology this March and will be helping with modeling human behavior for human-robot collaboration. Shuangqi Luo will continue to help us with probabilistic generative models for behavior modeling in robots and humans before applying for his Ph.D. in the United States.
  • 2017 June IROS Papers accepted
    We had two papers accepted for publication at IROS 2017 to be celebrated in Toronto Canada! Congratulations to all involved!
  • 2017 May ICRA Workshop Presentation
    We had a great time presenting our paper on Online Decision Making for Robot Contact Tasks at the AI in Automation Workshop at ICRA.
  • 2016 OCT Conference Proceeding
    Our work “Robot Contact Task State Estimation via Action Grammars. was accepted for publication at the 2016 IEEE-RAS International Conference on Humanoid Robots (Humanoids) on Nov. 15-17, 2016 in Cancun, Mexico.
  • 2016 JUNE Journal Proceeding
    Our paper: “Proposal of a Shape Adaptive Gripper for Robotic Assembly Tasks” was accepted for publication in the Journal of Advanced Robotics.
  • 2016 MAY Workshop
    Our Work “Task, and Skill Characterization for Contact Manipulation Tasks” was presented at the 2016 Robots and Science Conference (RSS): Bootstrapping Manipulation Skills in Ann Arbor, Michigan.
  • 2016 Mar ROS By Example (ROS入门实例) Now Available in ChineseRBX1 Now in Mandarin Chinese.
    ROS By Example in Chinese ROS By Example, the first book published on ROS, is now available in Chinese, thanks to the translation efforts of Juan Rojas, former Assistant Professor of Robotics at Sun Yat-sen University and currently a Research Associate Professor at Guangdong University of Technology, and the sponsorship of Jenssen Chang, Owner of Gaitech International Ltd., an innovative robotics solution provider based in Hong Kong, Seoul, Taipei and Shanghai, and an active promoter of ROS education in Asia. The new Mandarin translation can be obtained in print from Worldwide @ TaobaoWorld, DangDang.com, and JD.com.You can also seen the announcement in the official ROS page.5 reasons why you should by RBX Vol1. Hydro

     

    • The RBX series, written by Peter Goebel, enjoy the largest number of editions in the ROS literature. Patrick began writing ROS books when ROS issues its Electric version, and later, Fuerte, Groovy, Hydro, Indigo, and soon Jade. There is no other book with as many revisions. There is no other author with as much commitment.
    • RBX has two volumes: It’s the only ROS book in the market that covers as varied topics as the RBX series does.
    • It has the best code base and support. Both RBX1 and RBX2 volumes, have excellent code base on github. The code has been extensively tested and refined and covers from simple examples to very advanced ROS examples.
    • It has the best community support. Hundreds of users remain active only for this book in the Google groups @ https://groups.google.com/forum/#!forum/ros-by-example.
    • RBX by Example was translated by a team of Robot students. It was directed by Dr. Juan Rojas who has researched robotics for over 14 years and is a ROS expert along with a dedicated team of Robot engineers. You can be sure to get a great translation of the book!
  • 2016 Mar Guangdong Province Science and Technology Grant Approved:
    Grant obtained in cooperation with Han’s Motor and the Guangdong Institute of Automation to research 3C Production and Automation. Robot Production Line for 3C Industry Collaboration.2015 Guangdong Province Technology Projects. Special Fund
    Zhang Guoping, Rojas Juan, Zhou Xuefeng
    Category: The Forefront of Key Technological Innovation
    Dates:                    2015.08-2018.07
    Funding:               5,000,000RMB
    Project Number: 2016B0911006面向3C行业产线的协作机器人研发与产业化
    张国平, Rojas Juan, 周雪峰
    2015广东省科技计划项目,专项资金类别: 前沿与关键技术创新. 2015.08-2018.07
    五百万
    2016B0911006

2015 June Guangdong Province Science and Technology Grant
We got approved for a grant from the Guangdong Province Science and Technology Foundation in cooperation with SIAT, GDUT, and the robot manufacturer GSK!!Precise RobotPerception, Recognition and Integration of Complex Information Technology and its ApplicationsXia Zeyang, Guan Yisheng, Rojas Juan, Peng Jianqiong;
2014 Guangdong Province Technology Projects. Special Fund Category: The Forefront of Key Technology Innovation.
Dates:                 2015.06-2018.5
Funding:            3,000,000RMB
Project Number: 2014B090919002机器人精准感知、识别与复杂信息融合技术及其应用
泽洋, 管贻生, Rojas Juan, 彭健琼
2014广东省科技计划项目, 专项资金类别: 前沿与关键技术创新,
2015.06-2018.
3三百万.
2014B090919002.

2014 June Baxter Robot
Our lab was the first recipient of the Baxter Research Humanoid Robot in China!