After hours of testing, I'm thrilled to announce that you can now run the MIT Mini Cheetah in Gazebo using github.com/chvmp/champ. #goROS
Pinned Tweet
GIF
4
42
205
Follow
Juan Jimeno
@joemeno
Learning how to make machines walk.
Joined July 2009
Juan Jimeno’s Tweets
When you hired the wrong intern
Quote Tweet
It’s time for Atlas to pick up a new set of skills and get hands on.
1:10
21.2M views
7
my project is out!
MuJoCo MPC (MJPC) is an interactive tool for real-time behavior synthesis with predictive control algorithms.
paper: arxiv.org/abs/2212.00541
code: github.com/deepmind/mujoc
video: dpmd.ai/mjpc
13
156
823
Show this thread
How good of a BERT can one get in ONE DAY on ONE GPU?
With all the recent studies about scaling compute up, this paper takes a refreshing turn and does a deep dive into scaling down compute.
It's well written, stock full of insights. Here is my summary and my opinions.
🧶 1/N
38
619
3,025
Show this thread
Text-to-3D is around the corner...
I made this with Point-E on (link below 👇). Prompt: "Beautiful Christmas tree, red and green"
I spent thousands of hours making 3D models by scanning. Being able to generate 3D from text is... sorcery.
2023 will be WILD.
37
213
1,864
Show this thread
Creating photorealistic simulations of unstructured scenes is hard. Using NeRFs we turn 5min videos into simulations, train vision-guided policies for humanoid robots, and show zero-shot transfer to the real world!
abs: arxiv.org/abs/2210.04932
project: sites.google.com/corp/view/nerf
4
77
303
Thought-provocative new paper from : what if we could replace backpropagation with something better?
62
772
3,203
Show this thread
v0.1.12 Released 🎉
We have added initial mesh export functionality 🫖
#NeRF #nerfacto #neuralrendering #mesh
4
38
300
Show this thread
Enough sim for today
GIF
1
13
New 2023 goal - Finish CHAMP's ROS2 port before ROSCon 2023
1
13
Show this thread
3
6
36
#RSJ2022 で強化学習による歩様生成について先輩が発表します！SDK の High Level コマンドではなく、自前の歩様です。
4日目 9/8(木) PM1「多脚歩行ロボット」セッション、講演番号 4E2-03。
吉田 智章, 入江 清, 原 祥尭, 鈴木 太郎, 友納 正裕, "Sim-to-Real 転移を用いた四脚ロボットの歩容生成"
1
54
204
This demo uses an existing RVR robot package spawned with Linorobot’s bringup package. Once IMU and odom topics are matched, all Linorbot’s exisiting features can be used. Hopefully, users find this useful for integrating/prototyping new ROS2 based robots.
5
Show this thread
Finished writing a demo launch file for Linorobot’s new feature - Saddle. This allows arbitrary robots to piggyback on Linorobot’s existing packages, making it easier to integrate SLAM and Nav2 capabilities to newly built/non-ROS robots. github.com/linorobot/lino
6
10
53
Show this thread
My first blog post ever! Be harsh, but, you know, constructive.
Too much efficiency makes everything worse: overfitting and the strong version of Goodhart's law
sohl-dickstein.github.io/2022/11/06/str
🧵
35
216
926
Show this thread
Today, Boston Dynamics celebrates 30 years of innovation, exploration, and collaboration! We're excited to see what the next thirty years holds for us and for the robotics industry.
86
627
2,909
Gripping to see more language models used in robotics. Robots could only do much with explicitly programmed behaviors
Quote Tweet
How can robots perform a wide variety of novel tasks from natural language?
Execited to present Code as Policies - using language models to directly write robot policy code from language instructions.
See paper, colabs, blog, and demos at code-as-policies.github.io
long 
Show this thread
0:20
36.1K views
4
Announcing the 1st "Workshop on Pre-training Robot Learning" at , Dec 15.
Fantastic lineup of speakers: Jitendra Malik, Chelsea Finn, Joseph Lim, Kristen Graumen, Abhinav Gupta, Raia Hadsell.
Submit your 4-page extended abstract by September 28.
sites.google.com/view/corl2022-
3
35
114
New toy has finally arrived. Waited a year for this
2
49
Today we're launching the Farama Foundation, a new nonprofit dedicated to open source reinforcement learning, and we're beginning by maintaining and standardizing all the major open source reinforcement learning environments. Read more here:
15
118
394
New paper alert!
I spent 2021 building a framework for proprioceptive state estimation for all types of legged robots, and we report on some of our success. Work done with and .
Arxiv: arxiv.org/abs/2209.05644
3
20
110
Show this thread
My favorite lightning talk this year - BlenderProc2 - “A procedural Blender pipeline for photorealistic rendering” from github.com/DLR-RM/Blender
GIF
1
2
I could only imagine the complexity of the hardware but imagine PR2 variants being assembled/built from home #ROSCon202
1
1
18
If you were unable to join us in Kyoto for #ROSCon2022 we have two free live streams of the event available:
The first stream is available here:
vimeo.com/event/2520068/
The second stream is available here:
vimeo.com/event/2520087/
The show starts in about two hours!
1
19
26
これか。 github.com/ros-tooling/ro rosbag2のストレージプラグインなので、オプション指定するだけで、これまでのrosbag2の流儀のまま使える。これは使いたい。 #ROSCon2022
7
7
Show this thread
Do you use Xacro? Let me show you some alternatives at my #ROSCon2022 talk in a few hours😉!
Check out this repository to get started with more sophisticated templating engines and compare them to Xacro right now👇
5
29
1
5
18
Attending ROSCon next week? Renato Gasoto, Robotics/AI Engineer at NVIDIA, will be discussing simulation environments with Gazebo and ISAAC Sim running in parallel. Streaming available during the conference. Here's the content he'll cover in this session.
5
17
Excited to see this approach replace the more traditional object / instance detection approach. Helps predict samples from high dimensional distributions.
Quote Tweet
Introducing Pix2Seq-D, a generalist framework casting panoptic segmentation as a discrete data generation task conditioned on pixels. Works for both images and videos, with minimal task engineering.
arxiv.org/abs/2210.06366
work w/ Lala Li, @srbhsxn @geoffreyhinton @fleet_djShow this thread
0:04
22.3K views
13
27
270
Watching this CVPR talk on Occupancy Networks excites me so much. The amount of online perception for spatial reasoning is just nuts.
Quote Tweet
Presented some of the recent work from the Tesla Autopilot team at CVPR this year, especially about "Occupancy Networks" - our approach to solve general obstacle detection and using it to enable sophisticated collision avoidance. Full talk here: youtu.be/jPCV4GKX9Dw (1/12)
Show this thread
3
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.
GIF
4
28
178
Show this thread
Videos contain a wealth of information about the visual world, but self-supervised methods have yet to use them for learning good image representations. Introducing VITO, a new method which discovers semantic concepts by binding content across time: dpmd.ai/dm-vito 1/
GIF
7
66
316
Show this thread
For those asking in the comments, if you are curious about his research, this is an older book on the topic that he wrote
2
5
78
Show this thread
Today I had coffee with an MIT PhD who, in an effort to build AGI and mathematically prove free will, is coming up with an alternative to back propagation. How was your Monday?
56
48
894
Show this thread