SRA VJTI (@SRAVJTI) header
SRA VJTI (@SRAVJTI) profile picture

The Society of Robotics and Automation. As the name suggests, we deal with Robotics, Machine Vision and Embedded systems

Mumbai, India
Joined October 2013
Photo by SRA VJTI (@SRAVJTI)Photo by SRA VJTI (@SRAVJTI)Photo by SRA VJTI (@SRAVJTI)Photo by SRA VJTI (@SRAVJTI)Photo by SRA VJTI (@SRAVJTI)Photo by SRA VJTI (@SRAVJTI)Photo by SRA VJTI (@SRAVJTI)Photo by SRA VJTI (@SRAVJTI)Photo by SRA VJTI (@SRAVJTI)Photo by SRA VJTI (@SRAVJTI)
big driving models, tiny RC car, and a whole lot of experimenting. exactly the kind of stuff we love building at SRA
meet McQueen. an autonomous RC car built to test how pretrained driving models from @comma_ai perform in the real world, using a single RGB camera and remote inference.
24
533
pretty cool seeing YUMI come together! from a hacky handheld interface to collecting demos and deploying learned policies on a real YAM. powered on the robot side by our very own Karma :) github.com/SRA-VJTI/karma
introducing yumi a hacky $50* umi for the YAMs. collect data quickly -> visualize -> train & deploy *camera's not included - used from the other yam lol
23
470
open-source robotics is what gets us building
I always wondered what would have happened if the robots also had the sense of touch. So with that curiosity in mind, me and @mahi_zade0712 worked upon the "mission mimosa". result? policies converge faster and are sample efficient due to the flexi-tac on SO101.
2
27
685
built our very own intelligent robot control stack: github.com/SRA-VJTI/karma
we just oss'ed the robot control stack for achieving this - Meta Quest VR teleoperation - Single & dual-arm YAM/SO101 control - first class MolmoAct2 policy inference - Human-in-the-loop + DAgger data collection - data & rollout recording in @LeRobotHF v3 - based on the openpi-control interface from @physical_int github.com/sra-vjti/karma
20
628
If you are interested in learning how a slicing software actually works. My blog may help you ;) ashish4225.github.io/#/blog/…
1
6
1
46
5,225
e2e neural network cloth folding
cloth folding robot. with @tirth_gada
18
1,379
gpt6 tries wiping table
astra trying to wipe the table, a more sustained and intentional task testing its capabilities looking into the chain of thought explains and gives really good overview on the decision making. this is remarkable; slapping a flow head on this is robot AGI
16
737
checkout our coreless axial flux motor :D
no wires, no iron core, just copper spirals etched into PCB 4000rpm on the first proper run @BugejaCarl showed me its possible but mine runs on 3 phases and hope 🫡
1
1
29
1,012
What does it take to build the worlds cheapest lithography system @4rynv @hackerfabindia are building it using a BeagleBone Black paired with a miniature TI DLP board <$300 in optics.
8
31
17
228
44,854
A ~₹6k beamsplitter, ₹20k 10× objective and ₹4k bare camera sensor replace expensive proprietary optics. The enclosure? 3D printed.
1
4
43
2,355
The hard part: UV exposure. They’re designing a custom flex PCB with 2 LEDs: UV for exposure, visible light for focus verification. rebuilding a ₹2.5L setup from scratch.
1
3
43
1,311
yams go brrr more cool demos coming soon 🦾
Ever thought smooth policy inferences on your hardware set up is impossible? I ran bspline dp on a 200 episode ee vs joint control dataset to sketch the difference in the smoothness mapped 40 percent sr for joint control dp
20
577
physical atari in sra our robot learning team is so good @sahilsapage @LakshyaLalwani7 @4rynv
i must say it plays better than me
1
1
30
2,119
SRA VJTI retweeted
i must say it plays better than me
reinforcement learning on physical atari
7
5
2
159
10,543
🎮
reinforcement learning on physical atari
16
616
SRA VJTI retweeted
and it works on hardware, inference on a macbook m3 pro. also great at handling highly precise placement / drop thanks to dense patch features solves the task more efficiently that ACT and diffusion.
really cool idea! reproduced patch policy on Push-T tldr is "dense ViT patch features + block causal attention > global pooled features" dense features like VLA and really high frequency. planning to test it on real hardware for some precise manipulation
10
10
1
120
12,866
SRA VJTI retweeted
i love working on cool hardware:) here, letting a rL agent control the hardware to play atari games, the agent only sees the game through camera
5
7
2
58
5,708
Love seeing club members doing great work for fun. Super cool @spr4bhu
Great job by @spr4bhu for porting picorv32, a ultra-minimal RISC-V core to Shrike-lite! You can literally design your own custom CPU on this $4 board now. The example is officially merged into our repo!
1
16
743
Intelligent Robots everywhere 🤖🦾
MolmoAct2 kinda just works out of the box in an unseen environment and with unseen objects. I find it really crazy that I can just download a model from the internet and plug it into my robot and it can do some tasks at a limited capacity right away. this video is at 2x speed
5
37
3,118
We started the year with the basics: c programming, mechanical design ,embedded systems and computer vision The best way to conclude the year is with our legacy 3dof manipulator based on ROS2 : MARIO 200 First year students, 30 mentors and a weekend full of fun and learning
5
14
6
160
6,765
We teach everything that is required to build and control your very own first manipulator using Robot Operating System(ROS2) Python programming-->Simulations-->Microros The entire pipeline to see your FK IK codes come to life
1
16
457