Zhongyu Li @ZhongyuLi4 Twitter profile | Pikagi

Pikagi

Zhongyu Li

@ZhongyuLi4

854

Followers

314

Following

10

Media

285

Statuses

PhD Student doing robotics @UC Berkeley, BS @Zhejiang University.

https://t.co/RhREOYFWqh

Joined March 2021

Don't wanna be here? Send us removal request.

Pinned Tweet

@ZhongyuLi4

Zhongyu Li

4 months

Interested in making your bipedal robots to be athletes? We summarized our RL work to create robust & adaptive controllers for general bipedal skills. 400m-dash, running over terrains/against perturbations, targeted jumping, compliant walking, not a problem for bipeds now.🧵👇

15

90

441

Last Seen Profiles

@TheSmoothOne52

@ErlangerRx

@BoroAthletics

@joshidisha2

@velzarii

@Olympic_FC

@itsstiara

@ADHSsoftball

@lanig_harmo

@elu_art

@neel_beniwal

@VmaniakJ

@inoshkacho

@JanitaKutl34604

@e5Drnx4kYeSETGv

@CBPuertoVaras

@sumire911348556

@Ilovestays15

@chaothiqqq

@hanachigusa_ch

@MeathBeo

@UWSLeague2

@nwspeedway

@reindeereks

@McBaine146

@__aniiyahh__

@TMU_PR

@Adrikdraws

@MoglieMostra

@field111111111

@SOSsanidade

@eeisenhauer13

@Mikehoppy09

@TrishaBranan

@cnp_en

@oliwkaaa0215

@ZhongyuLi4

Zhongyu Li

1 year

Our goal is to push the limits of the agility of legged robots. Using a single RL-based multi-task policy, a biped Cassie learns to perform various challenging jumps, including standing long jump, high jump, and/or multi-axes jumps, with significant robustness, in the real world.

9

36

296

@ZhongyuLi4

Zhongyu Li

2 years

Can a quadruped perform agile and dynamic locomotion with fast and precise manipulation (like ball interception) skills? We create a robotic goalkeeper using Mini Cheetah with deep RL and it shows an 87.5% goal save ratio! More details can be found in our recent paper and video.

4

26

178

@ZhongyuLi4

Zhongyu Li

2 years

I think the coolest thing of GenLoco is that, if there is a new quadrupedal robot developed, we can just download the pre-trained model and use it as a baseline controller for the new robot without worrying about developing a new controller from scratch! Welcome to try it out!

@svlevine

Sergey Levine

2 years

Can we train a *single* policy that can control many different robots to walk? The idea behind GenLoco is to learn to control many different quadrupeds, including new ones not seen in training. Code Video

4

78

377

4

30

168

@ZhongyuLi4

Zhongyu Li

3 years

I'm very excited to share our recent work of using RL for a versatile and robust walking policy on a bipedal robot Cassie! The video is here: . The paper is at . Huge thanks to @whshkan @xbpeng4 @GlenBerseth @pabbeel @svlevine !

Tweet card media

Reinforcement Learning for Robust Parameterized Locomotion Control of...

Developing robust walking controllers for bipedal robots is a challenging endeavor. Traditional model-based loco-motion controllers require simplifying assum...

www.youtube.com

3

23

90

@ZhongyuLi4

Zhongyu Li

1 month

when you are training RL agents and what they learned

@AMAZlNGNATURE

Nature is Amazing ☘️

1 month

2 raccoons trying to open a door

1K

15K

114K

1

0

20

@ZhongyuLi4

Zhongyu Li

8 months

Super exciting to see more people using animation for legged robots! Back in 2020, my first paper was to use animation to draw expressive motions for Cassie. At that time, RL on real bipeds was still a question (we used traj opti)! Now, we also used anm for RL jumping on Cassie!

@BotJunkie

Evan Ackerman

8 months

This new robotic character from @Disney Research is ADORABLE, and also demonstrates a new pipeline for deploying expressive behaviors that are robust enough to work in the real world. We talk to the Disney team about how it all works: #IROS2023

6

73

295

1

0

19

@ZhongyuLi4

Zhongyu Li

2 months

It is very cool to nicely combine BC and RL for a set of challenging loco-manipulation tasks!

@ZhengmaoHe

Zhengmao He

2 months

Quadruped robots already have four manipulators (aka legs), do we really need extra arms for them? 🤔 Dive into our latest work: quadruped robots handle complex loco-manipulation tasks autonomously with just legs! 🐾

1

21

85

0

1

19

@ZhongyuLi4

Zhongyu Li

1 year

The takeaway is: the multi-task policy improves the robustness. By learning from various jumping tasks, the robot can generalize the learned tasks, and pick the best to recover from poor leaning pose or perturbation. Paper: Video:

Tweet card media

Robust and Versatile Bipedal Jumping Control through Multi-Task...

This work aims to push the limits of agility for bipedal robots by enabling a torque-controlled bipedal robot to perform robust and versatile dynamic jumps i...

www.youtube.com

1

2

14

@ZhongyuLi4

Zhongyu Li

4 months

We found that RL can bring robustness and adaptivity for dynamic & real-time bipedal control, but only with a proper formulation. Using a long history of robot I/O, complementing it with a short history, training it end-end, and adding task randomization, are the recipes.

1

0

12

@ZhongyuLi4

Zhongyu Li

4 months

We conducted extensive benchmark and ablation studies in both sim and real to support this. Details are in the paper: Summary video: This work cannot be done without @xbpeng4 @GlenBerseth @svlevine @pabbeel & Koushil

Tweet card media

Reinforcement Learning for Versatile, Dynamic, and Robust Bipedal...

This work presents a comprehensive study on using deep reinforcement learning (RL) to create dynamic locomotion controllers for bipedal robots. Going beyond ...

www.youtube.com

1

0

11

@ZhongyuLi4

Zhongyu Li

2 years

We decompose the goalkeeping task into 2 subtasks using RL: manipulation planning and locomotion control for multiple dynamic skills. This work is with X. Huang and a team of undergrads, and @xbpeng4 and K. Sreenath. Paper: Video:

Tweet card media

Creating a Dynamic Quadrupedal Robotic Goalkeeper with Reinforcement...

We present a reinforcement learning (RL) framework that enables quadrupedal robots to perform soccer goalkeeping tasks in the real world. Soccer goalkeeping ...

www.youtube.com

1

0

9

@ZhongyuLi4

Zhongyu Li

2 years

btw, a more general takeaway from GenLoco is, even when we are focusing on developing robot-specific controller using RL, randomizing robot morphology/kinematics can help to improve the robustness of the policy in sim/sim2real.

0

0

8

@ZhongyuLi4

Zhongyu Li

2 months

Tracking a reference motion🧐

@ShouldHaveCat

Why you should have a cat

2 months

Tweet media one

611

20K

243K

0

0

7

@ZhongyuLi4

Zhongyu Li

3 years

Not the first time using RL for locomotion control on bipeds but one of the first time using RL to directly output desired joint positions to control while showing great robustness! The cover figure is from our previous animation work though--this is another interesting paper😉

@techreview

MIT Technology Review

3 years

For the first time ever, a two-legged robot has taught itself to walk entirely using reinforcement learning.

6

70

191

0

1

6

@ZhongyuLi4

Zhongyu Li

1 year

This work cannot be realized without the help from @xbpeng4 @GlenBerseth @svlevine @pabbeel and Koushil.

0

0

4

@ZhongyuLi4

Zhongyu Li

8 months

@mengdibellaxu very cool work!

0

0

1

@ZhongyuLi4

Zhongyu Li

2 months

It is progressing soooo fast!!!

@tonyzzhao

Tony Z. Zhao

2 months

Introducing 𝐀𝐋𝐎𝐇𝐀 𝐔𝐧𝐥𝐞𝐚𝐬𝐡𝐞𝐝 🌋 - Pushing the boundaries of dexterity with low-cost robots and AI. @GoogleDeepMind Finally got to share some videos after a few months. Robots are fully autonomous filmed in one continuous shot. Enjoy!

56

333

2K

0

0

3

@ZhongyuLi4

Zhongyu Li

4 months

There are more cool videos👇 complete 400m dash: more running: more walking:

Tweet card media

Supplementary Walking Experiments - RL for Versatile, Dynamic, and...

Supplementary video for our work "Reinforcement Learning for Versatile, Dynamic, and Robust Bipedal Locomotion Control" [https://youtu.be/sQEnDbET75g]

www.youtube.com

0

1

3

@ZhongyuLi4

Zhongyu Li

2 years

Happy to share a recent RAL-IROS paper on using multiple quadrupeds to collaborate and tow a load using cables, which can be slack/taut, while navigating in cluttered spaces, fully autonomously and online. A cool demo video is here:

Tweet card media

Collaborative Navigation and Manipulation of a Cable-towed Load by...

This work tackles the problem of robots collaboratively towing a load with cables to a specified goal location while avoiding collisions in real time. The in...

www.youtube.com

0

1

3

@ZhongyuLi4

Zhongyu Li

2 years

Glad to share a recent paper that uses RL to enable a quadruped to precisely shoot a soccer ball to random targets. We first train a robust shooting controller then train and finetune a motion planner. This paper is in the IROS Best RoboCup Paper finalist this year!

@svlevine

Sergey Levine

2 years

A fun project on training quadrupedal robots to score a goal: w/ Yandong Ji, @ZhongyuLi4 , Yinan Sun, @xbpeng4 , @GlenBerseth , Koushil Sreenath

4

10

57

0

0

3

@ZhongyuLi4

Zhongyu Li

1 year

@DrJimFan Thanks! The paper is still in the submission so it is not opensourced. But we do have the plan to opensource the code after everything is wrapped up.

1

0

3

@ZhongyuLi4

Zhongyu Li

3 years

New experiment paper using quadrepeds to autonomously explore unknown environments with a skill of jumping through the obstacles that is not traversable by walking only.

@junzengx14

Jun Zeng

3 years

Glad to share our recent work including collaborators @ZhongyuLi4 and @ayush1292 about quadrupedal robots. The idea of this paper allows the robots to switch behaviors between walking around obstacles or jumping through them in order to reach its goal for a navigation problem.

0

0

6

0

0

2

@ZhongyuLi4

Zhongyu Li

3 years

Want to be freed from infeasibility problem when using CBFs? Try our soft-decay method!

@junzengx14

Jun Zeng

3 years

I want to share our recent work accepted to ACC about control barrier functions, where the hyperparameter is optimized to ensure point-wise feasibility. paper link: github repo: Hope this method saves people from excessive tuning...

1

0

4

0

0

2

@ZhongyuLi4

Zhongyu Li

6 months

@JameScottX Thanks for your interest! The freq is 33Hz, and yes, Cassie is a torque-controlled robot. Happy to chat more and you can find more details in our paper:

Tweet card media

Robust and Versatile Bipedal Jumping Control through Reinforcement Learning

This work aims to push the limits of agility for bipedal robots by enabling a torque-controlled bipedal robot to perform robust and versatile dynamic jumps in the real world. We present a...

0

0

1

@ZhongyuLi4

Zhongyu Li

5 months

So cool!

@OrdonezApraez

Daniel Felipe Ordoñez Apraez 🇨🇴

5 months

Exited so see interest in our recent work. Please visit our project website after the holidays, as we will release all code and examples.

0

6

14

0

1

1

@ZhongyuLi4

Zhongyu Li

2 years

@freddealmeida that's cool! May I know if there is a publication on it? We will definitely add your work to reference if we missed it.

0

0

1

@ZhongyuLi4

Zhongyu Li

2 years

Glad to share our recently submitted work on a bipedal robot Cassie autonomously navigating unknown environments with height constraints. The entire autonomy runs in real-time and real-world! Paper: Video:

Tweet card media

Vision-Aided Autonomous Navigation of Bipedal Robots in Height-Cons...

Navigating a large-scaled robot in unknown and cluttered height-constrained environments is challenging. Not only is a fast and reliable planning algorithm r...

www.youtube.com

2

1

1

@ZhongyuLi4

Zhongyu Li

1 year

@ZiqingShi Those long wires are for the safety of the robot & the nearby human (me😆), in case the robot falls over.

1

0

1