Profile avatar
roboticseabass.com
Robotics software engineer impostor supporting research in manipulation, planning, and learning -- open-source contributor and educator on the side -- mediocre runner, music nerd, anti-influencer/hype/grift -- 🇨🇱🇯🇲🤘 -- roboticseabass.com/about
163 posts 1,824 followers 241 following
Regular Contributor
Active Commenter

The next ROS Deliberation Community Group meeting is Monday, March 3, 2025. * Spotlight: BehaviorTree.CPP / PyTrees / TurtleBot behavior demos * Behavior Tree interpretability survey from Örebro University * Group discussion: Experiences using behavior trees docs.google.com/document/d/1...

Made it through a packed day at the University of Washington. Not only did my robotics colloquium talk go well (I think), but I got to meet a handful of faculty and students, see the robotic manipulation lab space, and not be freezing in Boston. Thanks to Sidd Srinivasa and @adlarkin.bsky.social !

Fresh article about the RAI Institute! spectrum-ieee-org.cdn.ampproject.org/c/s/spectrum...

Just by switching to using type hints, I've already found a ton of places where better design choices get rid of circular imports. I can't go back to untyped Python... This reminds me of the time I became a believer of unit tests after the first test suite I ever wrote uncovered 3 separate bugs.

I was bored enough tonight to see what it would take to enforce Python type hinting in one of my personal repos. This is gonna take a while...

Okay then.

Got around to watching the latest Karpathy video. Well worth watching for people who aren't super conversant about the LLM ongoings, like me. The RL section was a bit more hand-wavy than the others, but minor criticism of an otherwise amazing piece of educational content. youtu.be/7xTGNNLPyMI?...

I had never read this before, but it's (expectedly) so good. rodneybrooks.com/rodney-brook...

In preparation for the ROS 2 Deliberation Community Group meeting next Monday, I've put together a simple repository showing the basics of task planning using the AIPlan4EU Unified Planning library. Check it out! github.com/sea-bass/lea...

Great blog post that echoes a lot of my views on OSS, and what sometimes keeps me awake about participating in this ecosystem. Do give it a read.

Whoever thought it was a good idea on Linux to use Ctrl+Shift+C to copy from a console and Ctrl+C to kill the potentially long-running program in said console... is not my friend today.

Are you a ROS user and interested in automated planning, behavior abstractions, and high-level robot autonomy in general? I've recently started helping run the ROS Deliberation Community Group, which covers these topics, and our first meeting of 2025 is coming up. discourse.ros.org/t/ros-delibe...

For the new batch of people flooding in, if you're seeing this post there is some likelihood you may be interested in these starter packs. Open source roboticists and educators: bsky.app/starter-pack... TAMP Camp -- people working on task and motion planning: bsky.app/starter-pack...

Really cool repo based on PyTrees for scenario-based testing! github.com/intellabs/sc...

I know it's not healthy, but sometimes the urge arises to knock out some seriously impressive work just to silence the haters.

PyRoboSim 3.3.0 is now available! This is the first release where other contributors did most of the work, which is a wonderful thing to see happen. Plus, all these enhancements were also suggested via Git issue by another user. Full release notes: github.com/sea-bass/pyr...

If an alien civiization capable of communicating with us ever arrives, I would pay money to sit them down with someone trying to explain Kubernetes to them.

Super excited to have been invited to give a talk at the University of Washington Robotics Colloquium. It's on the 21st, not the 14th... but regardless, I'm gonna really have to step my game up given the others speaking at this. 😰

I don't know who cares (I do), but Pinocchio 3.x was finally released on PyPi on Jan 4! pypi.org/project/pin/...

One of the most Massachusetts things I've seen in a while.

I gave my first university talk last October and have another lined up this February. Based on the last talk, it's clear I need to make changes. It was about 80% background and 20% research, but almost all the questions were on the research. As I think about rebalancing the talk... any pointers?

As someone who uses their spare time writing educational content this a) is totally unsurprising b) still hurts. Unironically, one would have a better experience just typing the questions into ChatGPT. Good positive feedback loop. youtu.be/-opBifFfsMY?...

I've learned that my superpower is literally being too stupid to write complex software. Hey, if it works it works.