For Day 1, we showcased our latest development in MagiClaw at the IROS2024 Workshop on Environment Dynamics Matters. Received great feedback and support from the crowd and recieved the Excellence Practice Award. Confirmed our schedule for the competition tomorrow with Peter, joined the presentation by Prof. Hadaddin, and ran into U-Xuan and many colleageus on-site. Had a great meal at the China Restaurant and see you tomorrow at IROS2024 @ Abu Dhabi. | The Design and Learning Research Group is participating in this year’s IROS conference with a team of seven members led by Prof. Song Chaoyang and six lab members, including Dr. Guo Ning, Han Xudong, Sun Haoran, Wu Tianyu, Ge Sheng, and Xu Ronghan, for one competition and two workshops with three robots.
MagiClaw for Workshop on Environment Dynamics Matters
We sincerely thank the organizers of the IROS2024 Workshop on Environment Dynamics Matters: Embodied Navigation to Movable Objects for accepting our late extended abstract submission. We were excited to share our latest development on the DeepClaw toolkit and eager to collect feedback from our fellow researchers on improving our design before we formally roll out the system for everyone to use.
Between 2015 and 1017, Dr. Song and Dr. Wan developed the original version of the DeepClaw system at Monash University as an internal library for conducting benchmarking research and experiments on robotic manipulation. After moving to SUSTech in 2018, with help from Ge Sheng, Sun Haoran, Dr. Liu Xiaobo, and Dr. Guo Ning (both were first-year doctoral students back then), the DeepClaw package was broadened to include software interfaces for all robots available in the lab with several demos, including the Jigsaw Puzzle demo that won the ICARM2023 Best Paper Award Finalist. Later, we redesigned the DeepClaw system with aluminum extrusions for rapid reconfiguration and used it for teaching ME336 on Collaborative Robot Learning. Dr. Wang Haokun (an undergraduate student back then) developed DeepClaw 2.0 with Qiu Nuofan and Wang Teng by introducing a pair of modified grilling tongs with our Soft Polyhedral Network fingers and developed an open-sourced user interface for data collection. During COVID-19, the DeepClaw system underwent another redevelopment round by Dong Yujian, Xiao Yang, and Wei Jinqu. It became portable as a toolkit with an online interface to be better used for teaching ME336. This educational effort was recognized by UNESCO and won an international teaching award. The design iteration of the DeepClaw system goes hand in hand with our research on vision-based tactile sensing with the Soft Polyhedral Networks.
After using the DeepClaw Toolkit for teaching ME336 for three years and with several of our latest developments on vision-based tactile sensing, we felt now is probably a perfect time to evolve the DeepClaw system again. We are soft-launching MagiClaw during this workshop organized by Prof. Liu Huaping, Prof. David Hsu, Prof. Guo Di, Dr. Tao Kong, and Prof. Abhishek Gupta.
