C Zhang @ChongZitaZhang Twitter profile

Pinned Tweet

C Zhang

@ChongZitaZhang

14 days

Time to rest...

1

5

65

Last Seen Profiles

@adri60327893501

@Dippin_MINA

@guuurl7l

@irenerbae

@zlib280395

@polisbhoy23557

@adult311

@_ToxicDesire

@yourchemicalcat

@Pl_Maurice

@Archangel_HT

@youngwildera3

@cnT410vObY72Q

@12baegirl

@Saruak12

@NSPW

@SICKYDOESMMA

@unimaniu

@stwmaniax

@pengen_stw

@stwmaniax

@JimmyWinSV

@jandakembangstw

@3ncrypt3d__

@ShunichiroNKMR

@mazimenasumire

@JoelWBerry

@CNRSIdFSud

@ED_Maldives

@logoo2x

@BodoMdB

@DenisMajcenovic

@BAN923429666887

@AmbikQueen

@pengen_stw

@jandakembangstw

C Zhang

@ChongZitaZhang

1 month

Is this the world model @ylecun want -- motion prior + model-based RL -- seems quite an old thing but made "visual" 🤔

0

24

174

C Zhang

@ChongZitaZhang

1 month

Every time I want to convey insights through paper writing - I need to read this paper again 🫡 (since 2023)

6

16

165

C Zhang

@ChongZitaZhang

21 days

A more efficient parallel elastic actuator design

2

20

142

C Zhang

@ChongZitaZhang

10 months

@CVL_ETH A bit off the topic: never devalue yourself, no matter what others say or do on you, even when it is your advisor, the one you are emotionally dependent on, and the one you have intimate relationship with.

1

8

110

C Zhang

@ChongZitaZhang

7 months

finally made it public! We trained end2end model-free policies for risky terrains that were reserved for model-based methods. Super agile, no motion references. Paves our way to the ultimate solution of quadruped locomotion.

1

12

107

C Zhang

@ChongZitaZhang

1 month

For anyone interested We have significantly updated the ABS manuscript (now it's 18 pages and almost everything is explained): Also, code here Also, it's now an RSS accepted paper with full score reviews

GitHub - LeCAR-Lab/ABS

Contribute to LeCAR-Lab/ABS development by creating an account on GitHub.

github.com

AK

@_akhaliq

5 months

Agile But Safe Learning Collision-Free High-Speed Legged Locomotion Legged robots navigating cluttered environments must be jointly agile for efficient task execution and safe to avoid collisions with obstacles or humans. Existing studies either

28

160

829

0

11

80

C Zhang

@ChongZitaZhang

16 days

It's also inspired by @BostonDynamics 's awesome works. The key question is -- to which extent we need model-based methods for motion synthesis, when #RL may support robust and versatile control w/ efficient computation? Another work for this Question:

Learning Agile Locomotion on Risky Terrains

Quadruped robots have shown remarkable mobility on various terrains through reinforcement learning. Yet, in the presence of sparse footholds and risky terrains such as stepping stones and balance...

arxiv.org

Wenli Xiao

@_wenlixiao

16 days

🚨 Without Any Motion Priors, how to make humanoids do versatile parkour jumping🦘, clapping dance🤸, cliff traversal🧗, and box pick-and-move📦 with a unified RL framework? Introduce WoCoCo: 🧗 Whole-body humanoid Control with sequential Contacts 🎯Unified designs for minimal

8

40

192

1

8

47

C Zhang

@ChongZitaZhang

4 months

Things are getting hotter and I feel hard to catch every paper.. # of papers after 2023 is almost equal to the number before 2023, and many are on humanoids! More and more challenging problems conquered, IMO, starting by @zhaomingxie 's early work in 2018

GitHub - zita-ch/bipedal-robot-learning-collection: Collection of high-quality robo learning papers...

Collection of high-quality robo learning papers for bipedal robots. - zita-ch/bipedal-robot-learning-collection

github.com

1

12

46

C Zhang

@ChongZitaZhang

16 days

#RL for #humanoid control is good -- but HOW-TO enable humanoid to fulfill contact sequences like PARKOUR jumping by END-TO-END RL, w/o any motion prior, and apply adaptive contact forces? See our work WoCoCo! I also made a brief explanation video here:

WoCoCo (Introduction): Learning Whole-Body Humanoid Control with...

WoCoCo: Learning Whole-Body Humanoid Control with Sequential ContactsWebsite: https://lecar-lab.github.io/wococo/Paper: https://arxiv.org/abs/2406.06005Auth...

www.youtube.com

Wenli Xiao

@_wenlixiao

16 days

🚨 Without Any Motion Priors, how to make humanoids do versatile parkour jumping🦘, clapping dance🤸, cliff traversal🧗, and box pick-and-move📦 with a unified RL framework? Introduce WoCoCo: 🧗 Whole-body humanoid Control with sequential Contacts 🎯Unified designs for minimal

8

40

192

1

8

30

C Zhang

@ChongZitaZhang

1 month

Some lessons on Unitree H1 (and perhaps G1🔥) Good: 1 cheap 2 100% rigid body, com centered, friendly for RL walking 3 ~direct motor, accurate torque when small 4 large battery 5 large torque and velo limit -- Bad: 1. Huge impact - not very durable, imu-unfriendly (Cont'd)

3

2

28

C Zhang

@ChongZitaZhang

7 months

Stone giants and octopus monster are wild

AK

@_akhaliq

7 months

Make Pixels Dance: High-Dynamic Video Generation paper page: Creating high-dynamic videos such as motion-rich actions and sophisticated visual effects poses a significant challenge in the field of artificial intelligence. Unfortunately, current

5

86

389

3

4

25

C Zhang

@ChongZitaZhang

11 days

I make a comparison for the 3 recent RL-based H1 humanoid whole-body teleop+imit works: 1) H2O () and 2) OmniH2O () by @zhengyiluo @TairanHe99 @Xialin_He 3) Humanplus () by @zipengfu @qingqing_zhao_ @Qi_Wu577

1

6

27

C Zhang

@ChongZitaZhang

1 year

@zdeborova If you can't find experts for each paper, students are quite good choices. 1. Many ug (mostly 4y) have top conf pub. 2. As students, we are serious with reviewing. We know we are 'noobs', we respect papers. 3. We also do reviewing to build our taste in our first academic stages.

0

23

C Zhang

@ChongZitaZhang

1 month

People should try to close the conceptual gap between frontier researchers and general audience. This was something published 2y ago with techniques (mostly optimization without AI) matured at least 3y ago put together as a journal paper. But today's audience still find it new.

Massimo

@Rainmaker1973

1 month

This swarm of 10 bright blue drones lifts off in a bamboo forest in China, then swerves its way between cluttered branches, bushes and over uneven ground as it autonomously navigates the best flight path through the woods. [📹 Zhejiang University]

75

199

878

1

2

21

C Zhang

@ChongZitaZhang

1 month

A significant research shift is happening in robotics: from structured priors being efficient, to structured priors being bottlenecks.

2

1

22

C Zhang

@ChongZitaZhang

1 month

Ironically

1

3

20

C Zhang

@ChongZitaZhang

9 days

Trivial today. Triple backflip: Learning from reference (more likely) or have phase observation and crafted stage-wise rewards (I once did w/ SAC) Others: Heavy orientation rewards together with velocity tracking. Sufficient randomization. Multi-expert likely but not necessary.

Unitree

@UnitreeRobotics

9 days

Daily Training of Robots Driven by RL Segments of daily training for robots driven by reinforcement learning. Multiple tests done in advance for friendly service humans.😊 The training includes some extreme tests, please do not imitate. #AI #Unitree #AGI #EmbodiedIntelligence

109

382

2K

1

2

17

C Zhang

@ChongZitaZhang

8 days

Seems that there are no commercial legged robots with both powerful and agile upper body? Talos seems quite decent with 6kg payload at ~1m arm spread, and has torque transparency. But the motions are very slow.

1

2

16

C Zhang

@ChongZitaZhang

5 months

@Toukairinn_FUZZ Hi, For your interest, we report the peak torques and joint velocities in the paper (and power = torque * joint_vel). Compared to the commonly used trotting gait (the sota is named "rapid" here), ours have higher velocity but much lower energy consumption!

1

2

12

C Zhang

@ChongZitaZhang

16 days

Weird fact: I've somehow become the only person who helped all of existing projects of [human->humanoid]+[whole-body control]+[teleoperation->autonomy], though none of these are led by me: 1. H2O, as co-author 2. Omni H2O, as co-author 3. HumanPlus, minor help on sim2real

2

0

15

C Zhang

@ChongZitaZhang

1 month

Huge surprise for me to see this -- world model with robustness on humanoid.

Yen-Jen Wang

@wangyenjen

1 month

#RSS2024 Denoising World Model Learning (DWL) is accepted by RSS 2024 with all 4.0 scores! DWL is my first RSS paper ever, thanks to @JianyuChen_THU , Xinyang Gu from @roboterax , and Xiang Zhu, Chengming Shi, @GYanjiang , Yichen Liu from @Tsinghua_IIIS . We will publish this paper

3

9

64

1

2

14

C Zhang

@ChongZitaZhang

15 days

Fundamental knowledge of #robotics by @realyunlong

GitHub - yun-long/Foundations-of-Robotics: Learn fundamental knowledge in robotics

Learn fundamental knowledge in robotics . Contribute to yun-long/Foundations-of-Robotics development by creating an account on GitHub.

github.com

1

0

13

C Zhang

@ChongZitaZhang

3 months

Ten years later we will see in papers' introduction: "traditional RL-based controllers"

Chris Paxton

@chris_j_paxton

3 months

I've always thought this too. Although I'm mostly still an RL skeptic, I think the place this CLEARLY isn't true is walking robots; even Boston Dynamics has been getting in on the RL action these days

2

0

29

2

1

13

C Zhang

@ChongZitaZhang

2 months

solid papers from academia and open source RL infra put humanoids around the world on the same level. Some hyped companies seem to fall behind tho. But the fast progress will stop here - winner of next stage must have good state estimator and perception solution.

Maginative

@Maginative

2 months

The Beijing Humanoid Robot Innovation Center has unveiled "Tiangong," a full-size humanoid robot capable of running on electric power. Developed as a general-purpose humanoid robot platform, Tiangong stands 1.63 meters tall, weighs 43 kilograms, and can maintain a steady

0

4

19

0

1

13

C Zhang

@ChongZitaZhang

2 months

Welcome to our #ICRA2024 presentation at 10.30 am on 14 May. Room 301, award session for best paper on cognitive robotics (or robot+learning). #robotics #robotlearning My coauthor Jin will present our work. Project page:

Resilient Navigation

Resilient Legged Local Navigation: Learning to Traverse with Compromised Perception End-to-End Jin Jin* and Chong Zhang* , Jonas Frey, Nikita Rudin, Matías Mattamala, Cesar Cadena, Marco Hutter...

sites.google.com

C Zhang

@ChongZitaZhang

3 months

Are we ... in #icra finalist? (But why cognitive robotics for our navigation paper lol) Welcome to our session!

1

0

9

1

0

13

C Zhang

@ChongZitaZhang

2 months

Got 2/2 co-first-authored papers accepted by #RSS2024 . Thanks to all reviewers and ACs -- rly constructive and insightful comments. The rebuttal results are confusing tho. Seems that AC manually updated the reviews based on the reviewer feedback. 😂

1

0

13

C Zhang

@ChongZitaZhang

11 days

For @CMU_Robotics people: if you attend @corl_conf in Germany this year, you can: 1. travel in Schengen freely 2. go to Iceland, which is also in Schengen area 3. enjoy aurora, etc. 4. fly DIRECT from Iceland to Pittsburgh with 400 dollar.

0

12

C Zhang

@ChongZitaZhang

4 months

Test of time after 3 years

Antonin Raffin

@araffin2

3 years

Types of Reinforcement Learning Paper Original image: @xkcd

14

161

786

0

1

11

C Zhang

@ChongZitaZhang

5 months

@Toukairinn_FUZZ and here is another plot comparing the energy consumption for different gaits and controllers. I think, there is indeed another trade-off in agility vs energy👀

1

9

C Zhang

@ChongZitaZhang

1 month

It happens that I'm not found.

1

0

11

C Zhang

@ChongZitaZhang

7 months

For autonomous legged robots in the wild, local navigation with online SLAM is an essential function. However, the perception is not always ideal-- how to save in such cases? Our exploration: use e2e RL to react to perception failures! video:

Resilient Legged Local Navigation: Learning to Traverse with Compro...

Under review for: ICRA2024Paper: https://arxiv.org/pdf/2310.03581.pdfWebPage: https://bit.ly/45NBTuh Jin Jin* and Chong Zhang*, Jonas Frey, Nikita Rudin, Ma...

www.youtube.com

2

3

11

C Zhang

@ChongZitaZhang

29 days

Many robotic behaviors were called dynamic in 2019 and are called quasi-static in 2024.

1

0

10

C Zhang

@ChongZitaZhang

1 month

Add a link so I don't need to Google next time

Emergence of Maps in the Memories of Blind Navigation Agents

Animal navigation research posits that organisms build and maintain internal spatial representations, or maps, of their environment. We ask if machines -- specifically, artificial intelligence...

arxiv.org

0

10

C Zhang

@ChongZitaZhang

24 days

Today's people/investors are far more tolerant of the new era grandma bots and seem to have a weird belief that AI and data can make humanoids break the fundamental limits of hardware systems.

Christian Hubicki

@chubicki

24 days

ASIMO is essential on any list of all-time real-life robots. Honda unveiled ASIMO in 2000 after being developed in secret since the 1980s. It could walk, dance, handle objects, and ran shockingly fast — it held the fastest humanoid robot record (9 kph) for over a decade.

1

2

22

0

1

10

C Zhang

@ChongZitaZhang

3 months

Are we ... in #icra finalist? (But why cognitive robotics for our navigation paper lol) Welcome to our session!

C Zhang

@ChongZitaZhang

7 months

For autonomous legged robots in the wild, local navigation with online SLAM is an essential function. However, the perception is not always ideal-- how to save in such cases? Our exploration: use e2e RL to react to perception failures! video:

2

3

11

1

0

9

C Zhang

@ChongZitaZhang

4 months

Our works (Human to Humanoid and Agile But Safe) are featured by IEEE Spectrum Video Friday! @TairanHe99 @_wenlixiao Twice within a month as the cover videos -- thanks for sharing!

0

4

7

C Zhang

@ChongZitaZhang

1 month

The RL policy: win the DARPA subterranean challenge, hike on Alps without a fall, published on Science #Robotics 🥰 Also the same policy: falls dozens of times during hundreds of hours test in the wild 😨 And now we find it can fail under small noises and reasonable commands 🤔

Fan Shi

@fanshi_robot

1 month

RL controllers for quadruped locomotion are great but black-box. Are they safe enough? Our work in #RSS2024 reveals the risk of small noises and low-frequency velocity commands falling SOTA controller, which helped win DARPA SubT Challenge, even on the simplest flat terrain.

2

17

77

0

3

9

C Zhang

@ChongZitaZhang

3 months

Humanoid is an interesting topic and works on it can bring innovative techniques for nonlinear multi-contact system control -- you don't need to trust those commercial boosters for making it work in warehouse

Watney Robotics

@watneyrobotics

4 months

ADA regulations in the US ensure virtually all public spaces where robots might work *must* be accessible by wheelchairs. Through extensive industry research and customer interviews, we've identified only two environments where anthropomorphic humanoid robots have a strict

4

2

10

0

1

7

C Zhang

@ChongZitaZhang

2 months

LLMs won't enable this -- need more advanced perception & low-level control & fast reaction Language intelligence is not all you need. At least we still need motor intelligence.

Eric Jang

@ericjang11

2 months

Every day I am humbled by how smart and agile and persistent animals are

2

1

92

0

7

C Zhang

@ChongZitaZhang

4 months

Finally someone tried to roughly reimplement @ki_ki_ki1 's SR work on @UnitreeRobotics quadruped lol! Not thoroughly tho (didn't guide the encoders, DR not enough), but all in good ways (elev map recon and state estim. --there's no official api AFAIK)

2

0

7

C Zhang

@ChongZitaZhang

5 months

Accepted by #icra

C Zhang

@ChongZitaZhang

7 months

For autonomous legged robots in the wild, local navigation with online SLAM is an essential function. However, the perception is not always ideal-- how to save in such cases? Our exploration: use e2e RL to react to perception failures! video:

2

3

11

1

0

7

C Zhang

@ChongZitaZhang

7 days

stop hyping unitree and check what Jemin did Actuation Transparency is so FKING IMPORTANT for such [agile but precise] motions. It may be unnecessary for a commercial application, but necessary for top-tier control research. I feel the same during my WoCoCo and earlier works.

5

8

73

C Zhang

@ChongZitaZhang

11 days

Gonna travel around US next week and visit my friends. Currently in my plan: UT Austin Caltech Stanford anyone wants to catch me -- ping me and I will update my list

4

0

6

C Zhang

@ChongZitaZhang

21 days

Then what's in the future? More powerful generative models, more modalities for weakness identification, and more effective co-design evolution. And someone needs to put them together. The examples here are with legged robots, but the philosophy is above.

1

0

C Zhang

@ChongZitaZhang

25 days

mbrl + first order for legged locomotion learning: how contact reduces gradient accuracy and how to fix it. sim only though

Ignat Georgiev

@imgeorgiev

25 days

We derive a bound on this sample error and find that back-propagating through contact and long trajectories drastically reduces gradient accuracy. 🧵(3/4)

1

0

9

0

2

6

C Zhang

@ChongZitaZhang

3 months

I 100% underestimated it

Unitree

@UnitreeRobotics

3 months

Unitree H1 The World's First Full-size Motor Drive Humanoid Robot Flips on Ground. Unitree H1 Deep Reinforcement Learning In-place Flipping ! #Unitree #UnitreeRobotics #AI #Robotics #Humanoidrobots #Worldrecord #Flips #EmbodiedAI #ArtificialIntelligence #Technology #Innovation

68

414

2K

1

0

6

C Zhang

@ChongZitaZhang

1 month

@antoine_leeman Why care about data efficiency when it's just simulation

2

0

6

C Zhang

@ChongZitaZhang

3 months

@toudaihantei

0

6

C Zhang

@ChongZitaZhang

2 months

With a short trend, AI people mind: it's exp(t) Control people mind: it's 1-exp(-t) I still feel very strong limitations for biped platforms (also for humans) Moreover, hard problems still exist and are not solved by (even scaling of) cheap systems

Eric Jang

@ericjang11

2 months

By 2028 humanoids will be the primary form factor used by most robot labs. I believed this 3 years ago and believe it more than ever today

1

4

42

0

6

C Zhang

@ChongZitaZhang

7 months

@keenanisalive Simply take the derivative of the gravitational potential w.r.t. each block's orientation (see the whole system as an articulation where the joints are the "corners").

3

0

6

C Zhang

@ChongZitaZhang

2 months

#bostondynamics has a very fast-paced dev team. Even when they were only doing MPC - if you didn't feel I'd say you don't understand control. Now it's too easy for them to add some learning things. Everyone doing research on #humanoid must be clever about choosing the topic now.

0

6

C Zhang

@ChongZitaZhang

8 days

FUTURE ROBOTS SHOULD NOT BE WORSE THAN THESE ()

0

1

7

C Zhang

@ChongZitaZhang

8 months

@EugeneVinitsky Caught you!

2

0

6

C Zhang

@ChongZitaZhang

4 months

IMO, the most important point of such works is that, we can do whole-body behavior cloning for complex tasks in the future. #humanoid is important because they are anthropomorphic -- more than being able to locomote and manipulate

Tairan He

@TairanHe99

4 months

🤖 Introducing H2O (Human2HumanOid): - 🧠 An RL-based human-to-humanoid real-time whole-body teleoperation framework - 💃 Scalable retargeting and training using large human motion dataset - 🎥 With just an RGB camera, everyone can teleoperate a full-sized humanoid to perform

21

88

432

0

6

C Zhang

@ChongZitaZhang

15 days

… WoCoCo explained. One RL framework for #humanoid sequential contact tasks, e.g., parkour jumping, clap-tap dancing ... with task-agnostic sim2real. (Previously shared once, but this seems to create a YouTube preview friendly to twitter people)

WoCoCo (Introduction): Learning Whole-Body Humanoid Control with...

WoCoCo: Learning Whole-Body Humanoid Control with Sequential ContactsWebsite: https://lecar-lab.github.io/wococo/Paper: https://arxiv.org/abs/2406.06005Auth...

www.youtube.com

0

2

6

C Zhang

@ChongZitaZhang

30 days

@Stone_Tao @RoboDepot cool and close to being deployable. You may try to penalize (1) ground reaction force (2) foot velocity when contact (slippage) (3) foot acceleration (4) joint acceleration. Then the motion will be natural

1

0

5

C Zhang

@ChongZitaZhang

17 days

Water bird * Locomotion: Biped walk ✓ Swim ✓ Fly ✓ * Manipulation: With beak ✓ With wing ✓ With feet ✓ * Supported API Language ✓ Vision ✓ Tactile ✓ Smell ✓ Sound ✓ humanoid < water bird

2

1

5

C Zhang

@ChongZitaZhang

14 days

Wthhhh Such straight running while dragging the gantry. It says 3.5m/s but I estimate the peak velo to be around 4m/s. Also the gait is very stable-can be further accelerated and sustain long running.

Michael Overstreet #RobotJox

@i_bioloid

15 days

World’s fastest humanoid robot!

4

7

39

0

1

5

C Zhang

@ChongZitaZhang

1 month

Disclaimer: these are some personal feelings for some people curious about experience with unitree H1. It's only about the very specified one robot. Nothing about unitree support & others.

1

0

4

C Zhang

@ChongZitaZhang

3 months

@EugeneVinitsky In control we often say f(x,u) = Ax + Bu + w w is the noise -- so it looks quite reasonable to me and in robust control, we need to optimize u under w*=sup_{w} cost so in robust rl: optimize for largest noise can sometimes be even more effective (tho hard to explore)

1

0

5

C Zhang

@ChongZitaZhang

9 days

>>> and these AGI people successfully confused general people about robot and AI (Talk to random people that I work on robotics → "oh you study AI! Will they destroy the world?")

Chris Paxton

@chris_j_paxton

9 days

Robots don't need AGI to be useful in the real world, but AGI sure needs robots to exist in the first place

3

4

67

1

3

5

C Zhang

@ChongZitaZhang

1 month

2. Huge backlash (at least 2 deg on each joint of H1) 3. Limited versatility 4. No state estimator provided, no perception solution provided except two cheap sensors 5. Very inaccurate actuation when torque is high 6. Very small ankle torque compared to Cassie

2

0

5

C Zhang

@ChongZitaZhang

3 months

I once thought for good research you need to be serious. No. I began my robotics journey for fun, by my own, without any background or advisor. Building robots and making them move -- from building toys to building our life companions.

Kento Kawaharazuka / 河原塚健人

@KKawaharazuka

3 months

まさか, このロボットが国際会議に出るなんて.... 結構ちゃんと動いてます. 趣味と研究は紙一重！

1

240

953

0

5

C Zhang

@ChongZitaZhang

2 months

My Q would be how many of them have ever developed a velocity estimator, and how many have developed SLAM. Very few.

Jarett Gross

@JarettGross

2 months

Who else is excited for Humanoid Robots to take over? I found 34 companies building humanoid robots and made the most complete video featuring every one in existence on the internet! (for today at least) I spent over a month trying to find them all,

2

6

32

0

1

5

C Zhang

@ChongZitaZhang

1 month

Two recent papers from Nature and Science Robotics show that it's *good* (not just possible) to use event cameras/sensors for mobile robots: We need advanced perception for animal-level advanced sensorimotor skills.

Low-latency automotive vision with event cameras

Nature - Use of a 20 frames per second (fps) RGB camera plus an event camera can achieve the same latency as a 5,000-fps camera with the bandwidth of a 45-fps camera without compromising accuracy.

www.nature.com

0

5

C Zhang

@ChongZitaZhang

8 months

@Stone_Tao @EugeneVinitsky GPUs are better at rendering (ray tracing). Indeed the accuracy of dynamics is still bad on GPU. the contact forces can be arbitrarily huge on PhysX tho the motion looks "normal". And I don't feel mujoco-gpu does better. CPU+RLlib was a good solution (case: Cassie)

1

0

5

C Zhang

@ChongZitaZhang

4 months

I used to save my paper reading notes locally but I never open them again. So I'm considering putting them on twitter -> @RoboReading Reading notes of new papers will be shared and discussions are welcomed.

0

5

C Zhang

@ChongZitaZhang

9 days

@breadli428 Perceptive locomotion far from solved.

0

5

C Zhang

@ChongZitaZhang

3 months

like this line of work so much Object-oriented navigation Habitat simulator And control the dog Solid contribution, function in real world at a scale

Naoki Yokoyama

@naokiyokoyama0

5 months

Excited to share our latest work, Vision-Language Frontier Maps – a SOTA approach for semantic navigation in robotics. VLFM enables robots to navigate and find objects in novel environments using vision-language foundation models, zero-shot! Accepted to #ICRA2024 ! 🧵

1

40

207

0

3

5

C Zhang

@ChongZitaZhang

2 months

@CSProfKGD I would vote for yes if there are practical ways that can 1) clearly separate each author's contributions in big projects, 2) consider those failed trials and discussions into contributions 3) make it reader-friendly without overwhelmingly massive statements

0

5

C Zhang

@ChongZitaZhang

14 days

⚠️ How our humanoid robot got more and more tapes on it... #WoCoCo And maybe we need a savior controller! 🔄 Let's share more funny 🤖 videos like this by retweet!

Wenli Xiao

@_wenlixiao

14 days

🧠We really need safety guarantee for humanoid ‼️ Otherwise agile humanoid becomes fragile robot😭 We are really fascinated by recent humanoid demos in academia that trying to pushing the limit of - autonomous: HumanPlus, OmniH2O; - agility: Humanoid Parkour Learning, WoCoCo.

12

23

146

0

1

5

C Zhang

@ChongZitaZhang

22 days

@KyleMorgenstein I would care more about sim2real pipelines. There are no mature MJX solutions for sim2real (though Google released one at some initial stage, results are not convincing at all) -- then even it's 2x faster I wouldn't use it.

3

0

2

C Zhang

@ChongZitaZhang

21 days

Hmmmm

2

0

5

C Zhang

@ChongZitaZhang

22 days

@the_real_btaba @KyleMorgenstein Technically I want 1) procedure generation of random terrains 2) perceptive policy training with raw sensor obs 3) NN-based system ID from real to sim 4) neural state estimation from raw obs Many of these simply need a few lines of easy codes though.

2

0

4

C Zhang

@ChongZitaZhang

3 months

Added to my list tho. Rly hope my list can help people better contextualise the biped/humanoid rl things when writing.

GitHub - zita-ch/bipedal-robot-learning-collection: Collection of high-quality robo learning papers...

Collection of high-quality robo learning papers for bipedal robots. - zita-ch/bipedal-robot-learning-collection

github.com

C's Robotics Paper Notes

@RoboReading

3 months

Also they seem to cite their friends a lot while definitely downplaying significant contributions from labs such as OSUDRL.

1

0

4

C Zhang

@ChongZitaZhang

2 months

Teleoperation for manipulation is 🔥

Toru

@ToruO_O

2 months

Imitation learning works™ – but you need good data 🥹 How to get high-quality visuotactile demos from a bimanual robot with multifingered hands, and learn smooth policies? Check our new work “Learning Visuotactile Skills with Two Multifingered Hands”! 🙌

7

74

283

0

4

C Zhang

@ChongZitaZhang

1 year

↓this guy seems so confused

0

4

C Zhang

@ChongZitaZhang

7 months

@arthurallshire @EugeneVinitsky 😅 I experienced this three onboard PCs have three outputs for cmake

1

0

3

C Zhang

@ChongZitaZhang

11 days

Each of these works has pros and cons -- so we need to build upon them towards more perfect solutions. btw I do think Expressive Whole-Body Control by @xuxin_cheng is also a good work and close to the topic! also, PHC from @zhengyiluo is great!

2

0

3

C Zhang

@ChongZitaZhang

10 months

@_akhaliq Rly inspiring. But the tracking "flows" are on the task irrelevant background, which can make it challenging to generalize in the wild. There needs to be some "attention"

0

3

C Zhang

@ChongZitaZhang

23 days

you can manually add spring to your robot even when the company doesn't provide that

Quadrupedal Locomotion with Parallel Compliance: E-Go Design,...

{To promote the research in compliant quadrupedal locomotion, especially with parallel elasticity, we present Delft E-Go, which is an easily accessible quadr...

www.youtube.com

1

2

4

C Zhang

@ChongZitaZhang

4 months

@jan_drgona Then when will they find that requiring an accurate world model can be a fundamental limit xd

0

4

C Zhang

@ChongZitaZhang

5 months

@OlivierStasse Why not lol

1

0

4

C Zhang

@ChongZitaZhang

3 months

So hard to follow. today there are 100+ new #Robotic papers on arxiv .

0

4

C Zhang

@ChongZitaZhang

1 year

@kevin_zakka No worries. Others cannot cite your work if they cannot achieve something based on yours. LM things are far easier to implement than robotic things, but very few really “work” (e.g., in production). Hence, you can always expect you are cited by only high-quality impactful papers.

1

0

3

C Zhang

@ChongZitaZhang

30 days

Can tesla bot walk without Jerry Pratt's works published 20 years ago?😅 btw this year's Dynamic Walking is at IHMC.

Joseph 💎✌️🪑- Tesla Long Term Investor

@ShrimpTeslaLong

5 months

Figure CTO - Jerry Pratt. 20 years. Worked for "Institute for Human and Machine Cognition" and accomplished much? No. 11 years. Co-founded "Boardwalk Robotics". Did the bot get commercialized? No. 5 years. Interestingly, worked for a non-profit "Pensacola Mess Hall"? Chill

4

0

24

0

4

C Zhang

@ChongZitaZhang

2 months

@giffmana

0

4

C Zhang

@ChongZitaZhang

1 month

*Always work towards the ultimate solution or goal rather than rush for a temporary SOTA that makes no sense when later revolutionized by others.

0

4

C Zhang

@ChongZitaZhang

2 months

This is a bit far away from what I expect to be Parkour for humanoid (compared to atlas ofc), but definitely a good job and nice follow up of robotic parkour

Ziwen Zhuang

@ziwenzhuang_leo

2 months

Introducing 🤖🏃Humanoid Parkour Learning Using vision and proprioception, our humanoid can jump over hurdles, and platforms, leap over gaps, walk up/down stairs, and much more. 🖥️Check our website at 📺Stay tuned for more videos.

14

93

464

0

4

C Zhang

@ChongZitaZhang

1 month

the new issue is about advanced perception. So funny!

Contents | Science Robotics 9, 90

www.science.org

Science Robotics

@SciRobotics

1 month

The May issue of Science #Robotics is out! Taking inspiration from a robber fly’s eye, researchers developed a lens-free pinhole compound eye with a perovskite nanowire photodetector array. It exhibited a wide field of view and dynamic motion tracking.

3

21

88

0

4

C Zhang

@ChongZitaZhang

15 days

#RSS2024 program is out catch me here on July 17 around noon

0

4

C Zhang

@ChongZitaZhang

5 months

best collaboration with @TairanHe99 !

Tairan He

@TairanHe99

5 months

How to break locomotion agile-safe tradeoffs?Introduce Agile But Safe: Learning Collision-Free High-Speed Legged Locomotion: - Fully onboard - Agile (>3m/s) - Safe (collision-free guarantee) - Robust & versatile How? RL + model-free reach-avoid value! 👉

10

67

325

0

4

C Zhang

@ChongZitaZhang

2 months

@SoloGen @AMD AMD GPU is not graduate-student-friendly lol

1

0

4

C Zhang

@ChongZitaZhang

1 month

SRL's course at ETHz is definitely a funny one

T.Yamazaki

@ZappyZappy7

1 month

ワイヤ駆動多指ロボットハンドとセンサー付き人工皮膚強化学習で器用な操作をマスター #RoboticHand #artificial_skin #soft_skin #manipulation #3Dprinting #Education #STEM #robotics #ETHZurich #ETHZ #SRL #SoftRoboticsLab

0

28

134

0

1

4

C Zhang

@ChongZitaZhang

11 months

Do Machine Learning Models Memorize or Generalize? This is inspiring

Do Machine Learning Models Memorize or Generalize?

An interactive introduction to grokking and mechanistic interpretability.

pair.withgoogle.com

0

4

C Zhang

@ChongZitaZhang

7 months

@EugeneVinitsky Robotics sim2real struggles as no two real robots are created the same for reproduction xd

1

0

3

C Zhang

@ChongZitaZhang

1 month

Other cases: people doubt why limx dynamics build a wheeled-legged robot with biped mode (or think wheeled legged is a new design) while SwissMile already proves w/ demo for 2 years. Also, people don't understand why parallel simulation but it brought sota RL on legged in 2019.

1

0

3

C Zhang

@ChongZitaZhang

21 days

Funny as it is We can buy a $10000 camera for a Mars #robot , but only $100 for a street one. Same for special robots vs commercial ones. So it's vital to estimate the upper bound of reliability with a fixed budget before we build the robot. How? My discussion below🧵(1/n)

No Context Brits

@NoContextBrits

21 days

31

437

8K

1

0

4

C Zhang

@ChongZitaZhang

17 days

Guess we did make a breakthrough months ago. How fast should we reach as for a safety boudary? Consider transferring this to wheeled-legged robot + 3D locomotion and see what's gonna happen but len(todos) seems >10. 😅

0

1

4