
Yiting Chen
@YitingChen07
Followers
69
Following
28
Media
5
Statuses
21
PhD Student in Robot Manipulation @RiceUniversity
Joined April 2025
“You can’t make progress until you are able to measure it. Robotics still doesn’t have such a rallying call. No one agrees on anything.” I 💯 agree with the recent post from @DrJimFan. To break this impasse, we are excited to announce ManipulationNet ( https://t.co/DvoN2nvURq), a
9
23
108
This is the correct paradigm for benchmarking manipulation and now it arrives! We always welcome new task proposals; our infrastructure can support them all. More details at
As much as you can do system-level evaluation on the cheap this is the only way i think. - mail out standard object sets - have a client which records task performance and streams it
0
0
5
If your robot can’t tell colors, how can it perceive the world? If it can’t stably stack blocks, how can it understand physics? If it can’t link language to space, how can it reason? If it can’t contextualize tasks, how can it be generalizable? Before claiming “general reasoning
1
1
22
"One infrastructure to host a wide range of real-world benchmarking tasks" Now starts from the classic peg-in-hole assembly and block arrangement, to a general network of manipulation skills and abilities!
The pursuit of a unified benchmark for robot manipulation has been long-standing — and with ManipulationNet, it’s finally taking a leap forward! https://t.co/6ZnGzHykj2 ManipulationNet is a fully real-world, verifiable, and scalable benchmark for robotic manipulation. Anyone
0
0
4
A huge thank you to all our committee members: Kaiyu Hang, Kenneth Kimble, Edward H. Adelson, Tamim Asfour, @DanicaKragic, Xiang Li, @YunzhuLiYZ , Aaron Prather, Nancy Pollard, Maximo A. Roa-Garzon, Robert Seney, and @yukez, and our developer team Podshara Chanrungmaneekul,
0
0
3
ManipulationNet is born as a scalable infrastructure to support a wide range of real-world benchmarking tasks! Whether the tasks require interactive commands or are pre-configured for automated execution, they can all be benchmarked at scale within this paradigm. (6/7)
0
0
1
The mnet-server is a central manager for all client connections, which is responsible for handling all task-relevant activities. It supports task instructions delivery, performance logs collection in real-time for distributed benchmarking. In general, the mnet-server is
0
0
1
The mnet-client is a middle layer between the robotic system and the mnet-server (will be introduced in the following post) to support distributed manipulation benchmarking on standardized task setups. The robotic system communicates with the mnet-client through ROS services and
0
0
1
ManipulationNet provides a re-imagined paradigm, which relies on the integrated interaction between distributed hardware (for reproducible task setup in the real world) and software (for performance collection and task instructions delivery) to achieve the goal of distributed
0
0
1
ManipulationNet organizes benchmarking tasks in two tracks: 1) Physical Skills Track that evaluates manipulation skills in physical interaction-rich tasks; and 2) Embodied Reasoning Track that challenges robot capabilities in reasoning and multimodal grounding. (2/7)
0
0
1
Introduce XLeRobot: https://t.co/9hK0e8uNgC A practical low-cost household open source dual-arm mobile robot for general manipulation! Get yourself one with ~$660 and easily build it in <7hrs! (~$250 and <2hrs with a SO100 and a Lekiwi) Fully based on LeRobot More below:
9
65
393