Writing documentation for AI: best practices docs.kapa.ai/improving/wr...

My personal highlight of #ICRA2025 was my own presentation. I know its kinda shameless self promotion but this was my first experience and will be a lifetime memory! Conferences are these occasional pauses to receive necessary course corrections on your work!

Cool things happening at #ICRA2025 RoboRacers gearing up for their qualifiers

Wow, never really thought of it that way!

Next week I’ll presenting my work at #ICRA2025 in the Calibration session. Do visit to learn more about using IMMs as skidding estimators! TuCT12.3

Public education is the backbone of American democracy. In the richest country in the history of the world, we cannot abandon our teachers and our public schools.

I am looking for options to deploy VLMs/VLAs on edge devices. Came across this guy who deployed LLaVA on Pi. Are there better options? Especially for Xavier AGX / Orin kind of edge devices. youtu.be/0O8RHxpkcGc?...

Check out my paper at this year’s ICRA. TLDR: we are using adaptive estimation to guess robot model/environmental changes. arxiv.org/abs/2409.20554

This year looks to have the first fully autonomous, consumer home mobile manipulator, and it's not coming from Tesla or anyone: it's the Roborock Saros Z70, presumably coming for $2000 this summer to put your socks away: open.substack.com/pub/itcanthi...

A recent interview I did for the Brains and Machines podcast. www.eetimes.com/podcasts/rob...

An illustrated guide to never learning anything

AI tools for every human task would exist not because they are good or bad, but because people need an alternative to the effort of doing the task. AI tools for every human task would not matter, because people who enjoy doing the task for the sake of it would keep on doing it regardless. :)

I was wondering a while ago what happened to all the little delivery robot companies so I looked into it. I think it's so interesting that these things were everywhere in ~2018-2020, but then vanished! itcanthink.substack.com/p/what-happe...

More than the question on viability of the AV tech, this potentially is more about GM focusing on short term gains, which arguably are with ADAS.

IEEE RAS new journal Robotics and Automation Practice (RA-P) I believe this is going to be a pivotal step in bringing industry and academia a step closer! www.ieee-ras.org/publications...

Wow! Just wow. What would I give to be here! youtu.be/a0ANLUYyRYI?...

While I am extremely happy for all the ways NVIDIA is accelerating the robot learning processes, I don’t understand this need to constantly renew the nomenclature for vanilla concepts of modeling and control theory. www.nvidia.com/en-us/glossa...

Returning on the social after a while I am realizing that having an algorithm curate something like ‘for you’/‘discover’ could be useful to narrow down stuff. Especially you wanna catch up on 2-3 days worth of content :(

A significantly similar argument in engineering is between model based control (which is still very arts-y with coming up with basis functions, modeling,etc) vs using ML (IL/RL) for policy learning. Luckily the challenges with embodiment have kept the MBC relevant for now!

Probably cause the best test of robustness is Florida drivers 🤷🏻‍♂️

Key arguments from the ICRA debate “Generative AI will make a lot of traditional robotics approaches obsolete” (1/6) A thread 🧵 Have to say, just like my own mind, both the parties eventually seemed pretty on the fence and waiting for time to decide on this one. www.roboticsdebates.org

One of my fav times of the sem is being invited as guest judge in course project presentations. Evil smiles only 😈 JK, I am a kind judge 😇. Everybody passes!

I'd like to introduce what I've been working at @hellorobot.bsky.social: Stretch AI, a set of open-source tools for language-guided autonomy, exploration, navigation, and learning from demonstration. Check it out: github.com/hello-robot/... Thread ->

I am that guy, that guy is me :/ Stay away, please.

The more such tools are coming forth the less I wish to use them. Something about not taking passive aggressive suggestions from someone or *something* else. :(

Debate : Universities are irrelevant in an age of big tech and robotics A 🧵 thread on takeaways from this ICRA 24 debate -- paraphrased to best of my capabilities. (While being frank about my personal bias that this is not true :P) www.roboticsdebates.org

Once there was a solid community of researchers in control systems engineering whose writing seem funny, artistic, engaging and absolutely worth reading. Cometh are the days of template filling.

Happy Thanksgiving!

IMO its more about applicants’ confidence ( or a tad bit over confidence) in their capabilities. The more papers that I am writing, the more am I (over)estimating my skill to finish writing the article in a week. :(

Not sure how long I was on a lookout for this!

I am increasingly of the opinion that conference reviews should NOT be anonymous You should have to stand by your shitty ICLR reviews

The parable is really cool! — I identify as a technophile :p One a much serious note, designing task specific robot implementation >>> the “theory of everything “ like robot

I have seen some anti-peer-review takes on here, and, believe me, I understand the frustrations, but I would urge caution before doing or publicly saying anything too drastic. We live in age rife with misinformation and anti-intellectualism—not everyone wants academia to survive

Also every spring, but typical excuse is the pollen.

A fall dataset poem. Every fall, I think to myself, Wouldn’t it be great, if I made a fall dataset. Foliage that is amber and yellow, Will help the drone’s planning algo. Daydreaming on my lazy ass, I let the days of autumn pass. Every fall…

After a phenomenal couple of weeks of following great accounts, I am yet again overwhelmed by all the available content and hence in a dire need of bookmarking features/ optimized frameworks for ‘later’ reading.

This is so true. I see a lots and lots of young grad students trying reinvent the same old robust/adaptive control concepts while riding on the “trend” waves.

The fact that users relentlessly coming back on this platform despite occasional “server errors” is the social network equivalent of ignoring the red flags in the honeymoon of phase :p

Now that I have a decent lay of the land, perhaps a good time for a 'pin'-able tweet about me. I am a robotics researcher and a PhD candidate based out of Clemson Uni with research focused on field mobile robots -- with emphasis on combining learning & formal methods. ameyarsalvi.github.io

BlueSky or otherwise, there needs a push to preserve the “text primary” way of communication. Pictures/ videos are great but can be bit too overwhelming!

Is there like a walkthrough on how one could create their own “custom” feed?? It clearly seems like a hand feature.

I couldn’t attend Dr. Malik’s CoRL talk but I would like to believe UMich’s AI Symposium one would come pretty close. Unsure when the CoRL one will be available so mean time here is this : youtu.be/OlhhRFiygc0?...

Followed a lot of these accounts yet my feed is not “robotics” enough. Any ideas on how to get more of such content on your feed?

The one thing you can enjoy as a roboticist in grad school is the freedom to move on this spectrum: Technology <———> Theory One day writing cool code and next enjoy abstract math! #robotics #gradschool

RC - whattt now??? Teleoperated AVs #robotics

Rules of academia: Questions are more important than facts Answers are temporary Models are provisional Failure happens a lot Patience is a requirement When you’re lucky, hopefully you realize it Things don’t unfold linearly If there’s free food, get there early —Stuart Firestein

My first post here! Can’t be more grateful for all the starter packs! The beauty of open source in truest sense! Also is there a term for post on bluesky? Don’t wanna call it a “tweet”. Maybe “flutter” ???