Research Scientist at Boston Dynamics | Ph.D. CSE @uwcse | M.S Robotics Systems Development from @CMU_Robotics
Cambridge, MA
- Tweets 32
- Following 233
- Followers 508
- Likes 167
Mohak Bhardwaj retweeted
What if robot hands could acquire dexterity once, then reuse it to learn downstream tasks much faster?
Introducing ADEPT: a pre-training & post-training paradigm using RL entirely in sim, then zero-shot deploying visuo-tactile policies in the real world.
🔗adept-dexterity.github.io with Dex team @NVIDIAAI
8
33
258
66,718
Mohak Bhardwaj retweeted
Some more 'cherry' picked field test adventures with Spot! :)
If you visited the @uwcherryblossom, did you “spot” an unusual visitor among the blooms? Researchers in the @UW #UWAllen’s #Robotics group recently took advantage of some nice weather to take our @BostonDynamics robot dog for a stroll around campus. #AI 1/4
ALT A Boston Dynamics robot dog stands on a walkway through grass at the base of large cherry trees full of blooms
ALT Close-up of the chest and head of a Boston Dynamics robot dog framed by trees full of cherry blossoms
ALT Three smiling researchers dressed in casual clothing, one pulling a wagon, walk behind a Boston Dynamics robot dog with campus buildings and trees in the background
ALT A researcher holding a handheld controller follows a Boston Dynamics robot dog down a building ramp while another researcher standing a few meters away holding a laptop looks on
1
2
16
887
Mohak Bhardwaj retweeted
Boston Dynamics collaborated with NVIDIA to demonstrate DextrAH-RGB, a workflow for dexterous grasping from stereo RGB input.
The end-to-end policy for Atlas robot, trained entirely in NVIDIA Isaac Lab, transfers zero-shot from simulation to the real robot.
20
141
762
79,779
Mohak Bhardwaj retweeted
Our work with Boston Dynamics on RL training workflows for Atlas (upper torso) manipulation. An amazing team to work with!
Gina Fey, @mostlymohak @_mlutter, Alberto Rodriguez
@ritvik_singh9, Karl Van Wyk, @ankurhandos
There's a growing need in #robotics to shift from preprogrammed tasks to perceptive and adaptive robots for collaborative environments. 📈
Check out #NVIDIAResearch's work on robot dexterity, manipulation, and grasping, tackling adaptability and data scarcity. 👉 developer.nvidia.com/blog/r%…
2
16
101
7,276
Mohak Bhardwaj retweeted
Wishing you a holiday season full of light and laughter as we flip over into the new year!
85
411
2,248
195,757
Mohak Bhardwaj retweeted
Atlas is autonomously moving engine covers between supplier containers and a mobile sequencing dolly, using ML to detect and localize the environment fixtures and individual bin. There are no prescribed or teleoperated movements. piped.video/F_7IPm7f1vI
92
371
1,454
198,290
Come over to the poster session 4 at @l4dc_conf today if you're interested in learning more about our work on controlling a complex fluid dynamical system with RL directly in the real world.
proceedings.mlr.press/v242/b…
.@l4dc_conf is happening @UniofOxford this week!
Two fantastic papers coming up from our group @GoogleDeepMind in collaboration with @MIT!
First @timseyde continues the path of 'Q-learning is all you need' even for continuous control with 'Growing Q-Networks'
proceedings.mlr.press/v242/s…
Second @mostlymohak demonstrates RL for complex, real-world applications on the 'Box o’ Flows'
proceedings.mlr.press/v242/b…
9
1,101
Slightly belated life update: In April I started as a Research Scientist in the Atlas team at @BostonDynamics. Really excited to continue research in learning and manipulation with this amazing team and our shiny new humanoid!
13
14
500
47,534
Really excited to share that I successfully defended my PhD thesis at @uwcse earlier this month! My deepest gratitude to my committee and everyone who helped me along this 5 year journey!
4
1
40
3,937
Mohak Bhardwaj retweeted
Cooking in kitchens is fun. BUT doing it collaboratively with two robots is even more satisfying!
We introduce MOSAIC, a modular framework that coordinates multiple robots to closely collaborate and cook with humans via natural language interaction and a repository of skills.
4
41
182
26,439
Really excited to share work done during my @GoogleDeepMind internship. We introduce Box o' Flows, a benchtop control system for real-world fluid directed rigid body control and demonstrate how deep RL can be used to learn highly dynamic tasks directly on real hardware.
Reinforcement learning is most useful if a) demonstrations are hard to get and b) a system is hard to model.
( a) no imitation, b) no MPC etc)
Excited to share Mohak's internship report and the 'Box o Flows' enabling us to ask questions in this space!
arxiv.org/abs/2402.06102
1
3
22
4,995
A big thanks to my collaborators Jonas Buchli, Martin Riedmiller, @m_wulfmeier, Thomas Lampe, Michael Neunert, Francesco Romano, Abbas Abdolmaleki, Arunkumar Byravan and everyone else on the Controls team for their support.
378
Full version of our @NeurIPSConf 2023 paper titled Adversarial Model for Offline Reinforcement Learning is now available on arxiv arxiv.org/abs/2302.11048
Joint work with @tengyangx , @nanjiang_cs, Byron Boots and @chinganc_rl. Open source code coming soon!
8
37
3,078
Mohak Bhardwaj retweeted
Excited about sampling-based MPC but tired of assuming everything is a Gaussian and unsure about the right way to adapt the distribution online? Check out our #CoRL2022 paper on "Learning Sampling Distributions for Model Predictive Control"! (1/8)
Paper: arxiv.org/abs/2212.02587
1
20
98
Really excited to present our paper STORM: An Integrated Framework for Fast Joint Space MPC for Reactive Manipulation at @corl_conf! Checkout our oral talk on Thurs 10am GMT.
arxiv.org/abs/2104.13542
2
Mohak Bhardwaj retweeted
I have been working on a series that revisits core concepts in robotics in a contemporary light.
Starting with imitation learning and how it shows up in self-driving, on the simplest of problems!
piped.video/GDmhrAHxgQE
1
1
20
Mohak Bhardwaj retweeted
We would all like our robots to live and work alongside fellow humans. But how should robots imitate us?
In this 10-part series, I dive deep into imitation learning, and build up a general framework. A journey through feedback, interventions and more!🧵bit.ly/3Bxv80m
1/
1
11
42
Mohak Bhardwaj retweeted
Announcing another exciting #CoRL2021 tutorial: "Machine Learning Perspectives on Model Predictive Control" from Byron Boots of #UWAllen @uwcse @uw
Reminder- submissions close June 18th: robot-learning.org
#robots #learning #machinelearning #robotics #conference #robot
1
20
52
Excited to announce that our paper "Blending MPC and Value Function Approximation for Efficient Reinforcement Learning" recently got accepted to #ICLR2021!
arxiv.org/abs/2012.05909
Thanks to my co-authors Sanjiban Choudhary @sanjibac and Byron Boots for making this work possible.
1
6