Robotics papers

CMU Robotics Institute Videos from the Carnegie Mellon University's Robotics Institute. The RI is part of the Carnegie Mellon University School of Computer Science.

Heads up‼️ RI seminars are back underway for the spring semester! 🌷🤖

You can find the latest recorded talks from Andrea Bajcsy and Aaron Johnson on the CMU RI YouTube channel. Keep an eye out for more! ⬇️ #CMURobotics

www.youtube.com/@cmurobotics

Just subscribed to Le Chat Pro from @MistralAI
It has pretty much the same features as ChatGPT and it feels good to support 🇪🇺
We need a diversity of solutions to accelerate innovation!

"Our study reveals [...] increased model accuracy only marginally improves sample efficiency compared to learned models with identical horizons"

Surely an involved read from @tuda.bsky.social & @hessianai.bsky.social, but recommended if you are into model-based RL.

Also, ✨nice figures✨

Diminishing Return of Value Expansion Methods Model-based reinforcement learning aims to increase sample efficiency, but the accuracy of dynamics models and the resulting compounding errors are often seen as key limitations. This paper empiricall...

Diminishing Return of Value Expansion Methods

Authors: Daniel Palenicek, Michael Lutter, João Carvalho, Daniel Dennert, Faran Ahmad, and Jan Peters Fellow

pre-print -> arxiv.org/abs/2412.20537

#rl #reinforcement_learning #modelbased_rl #value_expension

For other 3D vision newcomers to blue sky: I highly recommend joning @chrisoffner3d.bsky.social 's list to follow the right people ;): go.bsky.app/Cfm9XFe

🎊 Happy new year robotics community !

Best wishes for the growing robotics community 🥂

Paper curation will resume tomorrow, but user lists and the starter pack are continuously updated.

Most notably, Pr. (@yiannisdemiris.bsky.social) from Imperial College joined us recently and has been added !

Release of open source software (inc #ROS packages) to control the Boston Dynamics Spot with #Hololens #AR glasses. Work by Dr Rodrigo Chacon at the Personal #Robotics Lab at Imperial College London. Papers with extensive user evaluations, code and more videos at github.com/ImperialColl...

Inspired by @wimlds.bsky.social , I looked for a "Women in Machine Learning" starter pack and couldn't find one.
So I created one! May have some mistakes. I'll try to do an AI ethics one next. 🤗
go.bsky.app/LT6CwNN

If you're training models with more than one loss term, I can again strongly recommend our ConFIG optimizer: tum-pbs.github.io/ConFIG/ , simply swap out Adam&Co. for ConFIG, and you can potentially see substantial reductions in your training loss 😁 We'd also be curious to hear how it works for you

A fresh approach to optimal control leveraging ideas from cognitive science.
With only 10 pages it is a recommended read for its original approach and the potential in further developments.

Overview of neo-free

Evaluation of neo-free

Neo-FREE: Policy Composition Through Thousand Brains And Free Energy Optimization

Authors: Francesca Rossi, Émiland Garrabé, Giovanni Russo

pre-print -> arxiv.org/abs/2412.06636

code -> github.com/GIOVRUSSO/Co...

#robotics #control #optimal_control #movement_primitives

We now have a starter pack for ROBOTICS. This should be a fun group.

Are you building robots or sharing thoughts/research on the state of robotics? 🤖🦿

Request to join if you would like to be added, before it fills up.

go.bsky.app/61LAgEj

Hi, I’d love to be added if you don’t mind !

The first is full but still incomplete. Here is the second one, also largely incomplete. Just say 👋 and you will be added.

go.bsky.app/DWo6RbA

We’ve been investigating how sim, while wrong, can be useful for real-world robotic RL! In our #NeurIPS2024 work, we theoretically showed how naive sim2real transfer can be inefficient, but if you *learn to explore* in sim, this transfers to the real world! We show this works on real robots! 🧵(1/6)

Point cloud segmentation is a crucial challenge in robotics, which this paper tackles without deep learning but by introducing a new member to the ever-growing families of positional embeddings.

While confusing at times, it is an interesting read if you are into machine learning for geometric tasks

🤖 "Robot" usually evokes a #quadruped or #humanoid system. However, they can also be... liquid ferrofluid marbles.

📄 This 100% 'machine learning' free paper is a sobering return to physics and reminds us that #robotics is, before all, the art of controlling motion.

Setup of Novel Magnetic Actuation Strategies for Precise Ferrofluid Marble Manipulation in Magnetic Digital Microfluidics: Position Control and Applications

Experiment from Novel Magnetic Actuation Strategies for Precise Ferrofluid Marble Manipulation in Magnetic Digital Microfluidics: Position Control and Applications

Novel Magnetic Actuation Strategies for Precise Ferrofluid Marble Manipulation in Magnetic Digital Microfluidics: Position Control and Applications

Authors: M. H. Sarkhosh, H. N. Pishkenari et al.

pre-print -> arxiv.org/abs/2412.02859

#robotics #control #manipulation #microfluidics

Overview of Point-GN: A Non-Parametric Network Using Gaussian Positional Encoding for Point Cloud Classification

Architecture of Point-GN: A Non-Parametric Network Using Gaussian Positional Encoding for Point Cloud Classification

Inference for Point-GN: A Non-Parametric Network Using Gaussian Positional Encoding for Point Cloud Classification

Point-GN: A Non-Parametric Network Using Gaussian Positional Encoding for Point Cloud Classification

Authors: Marzieh Mohammadi, Amir Salarpour

pre-print -> arxiv.org/abs/2412.03056
code (to come) -> github.com/asalarpour/P...

#point_cloud #classification #positional_embedding #non_parametric

Today we get two (don't get used to it) little gems from Iran to remind us all that robotics is not all deep and reinforcement learning !

#robotics #papers

I'm excited about scaling up robot learning! We’ve been scaling up data gen with RL in realistic sims generated from crowdsourced videos. Enables data collection far more cheaply than real world teleop. Importantly, data becomes *cheaper* with more environments and transfers to real robots! 🧵 (1/N)

From Instantaneous to Predictive Control: A More Intuitive and Tunable MPC Formulation for Robot Manipulators Model predictive control (MPC) has become increasingly popular for the control of robot manipulators due to its improved performance compared to instantaneous control approaches. However, tuning these...

From Instantaneous to Predictive Control: A More Intuitive and Tunable MPC Formulation for Robot Manipulators

Authors: Johan Ubbink, Ruan Viljoen, Erwin Aertbeliën, Wilm Decré, Joris De Schutter

pre-print -> arxiv.org/abs/2412.015...

#robotics #motion_control #mpc

Check out my #robotics lists and please RT to reach people I might have missed.

Robotics & ML -> bsky.app/profile/did:...

Robotics Professors -> bsky.app/profile/did:...

Starter pack will be updated to mirror Robotics & ML posting

In my continued effort to give visibility to researchers and lower the entry bar for robotics enthusiasts on 🦋, I have found that lists are a great tools to order content.

Lists can be pinned to the home screen, effectively acting as curated feeds.

With 20 spaces left in the #ResearchIntegrity Starter Pack, could we fill these with early career researchers/professionals or others who might not feel confident putting themselves on the list?

If you have suggestions for people who'd be a good fit, please let me know!

go.bsky.app/5NJ9Z4N

I made an HRI Starter Pack!

If you are a Human-Robot Interaction or Social Robotics researcher and I missed you while scrolling through bsky's suggestions, just ping me and I'll add ya.

go.bsky.app/CsnNn3s

Done !

GARField: Addressing the visual Sim-to-Real gap in garment manipulation with mesh-attached radiance fields While humans intuitively manipulate garments and other textile items swiftly and accurately, it is a significant challenge for robots. A factor crucial to human performance is the ability to imagine, ...

GARField -> arxiv.org/abs/2410.05038

RL-GSBridge: 3D Gaussian Splatting Based Real2Sim2Real Method for Robotic Manipulation Learning Sim-to-Real refers to the process of transferring policies learned in simulation to the real world, which is crucial for achieving practical robotics applications. However, recent Sim2real methods eit...

RL-GSBridge -> arxiv.org/abs/2409.20291

RoboGSim: A Real2Sim2Real Robotic Gaussian Splatting Simulator Efficient acquisition of real-world embodied data has been increasingly critical. However, large-scale demonstrations captured by remote operation tend to take extremely high costs and fail to scale u...

RoboGSim -> arxiv.org/abs/2411.11839