Computer Science Professor at Northeastern. Robotic manipulation, learning, perception, planning, control, and other stuff. My group: @HelpingHandsLab
Joined April 2022
- Tweets74
- Following157
- Followers755
- Likes1
I'm excited about this work. The equivariant model significantly reduces the number of demos needed to train good diffusion policies.
Not only is this an interesting approach, but it works great! Outperforms everything we benchmarked against.
Check out our new grasping work, OrbitGrasp! It achieved a 98% grasping success rate by learning the SE(3)-equivariant grasp quality distribution using spherical harmonics. I’ll be at #CoRL2024 to present this paper.
Instead of inferring a desired object pose directly, this method "imagines" a reconstruction of the entire scene in the target pose. Surprisingly, we find that this improves sample efficiency, even though we are inferring more information.
Robert Platt retweeted
#ICLR24
We proposed FourTran, a very sample-efficient 3D manipulation pick-place model.
1. It can learn a nontrivial 3D policy with less than 10 demos.
2. It represents 3D action distribution in Fourier Space.
Check it in the Poster Session 4 at 4:30 PM Vienna time (10:30 EDT)
Robert Platt retweeted
Transporter network is great for robotic pick-place, but how to use it in 3D where rotating the kernel for entire SO(3) is infeasible? Our work performs cross correlation in Fourier space to facilitate reasoning over SO(3). Check out at #ICLR2024, @RobinSFWalters is presenting!
#ICLR24
We proposed FourTran, a very sample-efficient 3D manipulation pick-place model.
1. It can learn a nontrivial 3D policy with less than 10 demos.
2. It represents 3D action distribution in Fourier Space.
Check it in the Poster Session 4 at 4:30 PM Vienna time (10:30 EDT)
Robert Platt retweeted
Registration is now open for Boston Symmetry Day on April 7th! We have an exciting lineup of speakers (@mario1geiger, @tesssmidt S Gortler, E Grigsby, @RobotPlatt, @_onionesque) + more to come.
Sign up here: forms.gle/8gZUh1BRL2FFatD58
Robert Platt retweeted
Hello world! The Boston Symmetry Group (pun intended) organizes events to foster community amongst Boston-area researchers and practitioners working on symmetry + machine learning.
Organizers: @StefanieJegelka @HLawrenceCS @dereklim_lzh @park_jungy @RobinSFWalters
Robert Platt retweeted
Congrats to my PhD student Sammie Katt for passing his defense titled "Bayesian Partial Observable Reinforcement Learning." Very cool work, Sammie! (with @faoliehoek, @RobotPlatt, and Lawson Wong).
Robert Platt retweeted
Interested in partial observability? Come to my online poster session at @CoRL2022 (online session 2, Sunday, Dec 18, 7:30-8.00 AM, Auckland time, i.e., Saturday, Dec 17, 1:30 - 2:00 PM, EST time) w/ @RobotPlatt @cjdamato @DianWang1007 @AndreaBaisero @LLPR_NEU @HelpingHandsLab.
Glad to share our #CoRL22 paper arxiv.org/abs/2211.01991 about using MDP solutions to efficiently learn POMDPs during offline training. We proposed a SAC-like agent that balances between acting like an MDP expert and for environment rewards.
Robert Platt retweeted
Excited to share our new SOTA grasp method in the Sim-to-Real Robot Learning Workshop at #CoRL2022. I will present the poster tomorrow 13:30-14 and 14:30-15:15 at Eng 405-460. Our work proposes a novel grasp representation as the graph edge and an SE(3)-equivariant model. 1/
Robert Platt retweeted
Very excited to share our work in the #CoRL2022 Workshop on Sim-to-Real Robot Learning! Our work leverages a novel simulation-based 3D data augmentation method and equivariant networks, helping achieve high data efficiency in imitation learning for robotic manipulation. 1/n
Robert Platt retweeted
This work is led by @saulbadman12 @DianWang1007, with @GuanangS, @_davidklee, @XupengZ, @RobinSFWalters, @RobotPlatt, @HelpingHandsLab. The full version paper: arxiv.org/pdf/2211.00194. Full video: youtube.com/watch?v=UHUmq-cO….
Robert Platt retweeted
I'll be presenting our #NeurIPS poster tomorrow on Wed 11-1am, poster session 3. Please come by and check it out!
How do you robustly imitate expert demonstrations, without access to the environment?
We present "Robust Imitation of a Few Demonstrations with a Backwards Model", to appear at #NeurIPS2022.
Paper: arxiv.org/abs/2210.09337
Code: github.com/jypark0/bmil
(1/4)
Robert Platt retweeted
We have 2 papers in the main conference #NeurIPS2022.
Poster: Wed 5-7
Shield Decentralization for Safe Multi-Agent Reinforcement Learning
neurips.cc/virtual/2022/post…
Poster: Thurs 5-7
Asynchronous Actor-Critic for Multi-Agent Reinforcement Learning:
neurips.cc/virtual/2022/post…
Looking for an easy-to-read review paper on Grasp Learning? I just wrote one for Annual Reviews that I'm excited about. Check it out:
arxiv.org/abs/2211.04895
I'm looking for PhD students for Fall 2023 to work at the intersection of Robotic Manipulation, Imitation Learning, and RL. I'm particularly interested in applying novel neural models like geometric DL. Apply by 12/15. @KhouryCollege @HelpingHandsLab
Robert Platt retweeted
We tested on 6 robot domains and transferred one learned policy to real hardware. Project website: sites.google.com/view/cosil-…. Joint work w/ Andrea Baisero @DianWang1007 @cjdamato
@RobotPlatt at @LLPR_NEU @HelpingHandsLab
Robert Platt retweeted
I am looking for a PhD student to start in Fall 2023 at @KhouryCollege focusing on the direction of 3D vision. Applicants from underrepresented groups are particularly welcome, where the college provides two-year fellowship support (domestic students only). Deadline is Dec 15.