We develop perception, control, & planning algorithms for robot autonomy | @CMU_Robotics | https://teapaw.org/t.co/gWjGiUaBeP | https://teapaw.org/t.co/9wO6amxfFc
Pittsburgh, PA
Joined June 2014
- Tweets537
- Following138
- Followers3.5K
- Likes461
AirLab retweeted
๐ค ๐๐๐๐: ๐๐๐ฅ๐-๐๐ฎ๐ฉ๐๐ซ๐ฏ๐ข๐ฌ๐๐ ๐๐ง๐๐ซ๐ญ๐ข๐๐ฅ ๐๐๐จ๐ฆ๐๐ญ๐ซ๐ฒ
Since the first learned inertial odometry 10 years ago, every method has needed ground-truth labels.
It doesn't have to: SSIO learns from raw IMU, fully self-supervised ๐งต๐
ssio-imu.github.io
Weโve witnessed the evolution of learned inertial odometryโfrom impractical to deployable, and from deployable to accurate.
The next frontier: from specific to generalizable state estimation.
๐ค ๐๐๐๐: ๐๐๐ฅ๐-๐๐ฎ๐ฉ๐๐ซ๐ฏ๐ข๐ฌ๐๐ ๐๐ง๐๐ซ๐ญ๐ข๐๐ฅ ๐๐๐จ๐ฆ๐๐ญ๐ซ๐ฒ
Since the first learned inertial odometry 10 years ago, every method has needed ground-truth labels.
It doesn't have to: SSIO learns from raw IMU, fully self-supervised ๐งต๐
ssio-imu.github.io
We donโt test with crash dummies. We test with Micah, at 10m/s, fully autonomous, off-road.
His body made it through every corner. His soul made it through most of them.
Check out our work on learned perception-conditioned dynamics models for high-speed off-road driving ๐
How your robot moves depends on what itโs driving over, especially when moving at >10m/s over loose gravel or ice ๐๐จ
What if dynamics models could be conditioned on exactly the terrain they expect to drive over? Introducing SpatialBEV!
Accepted at #CoRL2026 ๐ค
0/7
AirLab retweeted
How your robot moves depends on what itโs driving over, especially when moving at >10m/s over loose gravel or ice ๐๐จ
What if dynamics models could be conditioned on exactly the terrain they expect to drive over? Introducing SpatialBEV!
Accepted at #CoRL2026 ๐ค
0/7
AirLab retweeted
Failures are inevitable for robots that are deployed in the *big* world. We introduce ReSYNC, a system that leverages large-scale RL sim2real to recover from local failures and concept discovery to explain/avoid new failures. Check it out here: jaraxxus-me.github.io/ReSYNCโฆ
AirLab retweeted
MAC-Iยฒ learns how much to trust vision and IMU.
Metrics-aware covariance for both modalities enables robust VI fusion. We apply it to initialization and calibration in challenging scenes and motions.
mac-i2.github.io/
1/9
AirLab retweeted
Where do foundational models actually meet the real world?
That's the question behind the 3rd AI Meets Autonomy workshop at IROS 2026!
๐ Pittsburgh | ๐๏ธ Thu, Oct 1 | 8:00โ12:30 AM #IROS2026 ๐งต๐
AirLab retweeted
Excited to present our work "SuperMap: A Spatio-Temporal SLAM System for Visual-Language Navigation" at RSS 2026!
SuperMap is a living spatial memory for embodied AI โ it perceives the world, remembers its evolution, and supports reasoning and action.
Perceive โ Remember โ Reason โ Act
๐ Project Page: superodometry.com/supermap
#RSS2026 #Robotics #SLAM #EmbodiedAI #SpatialAI #CMURobotics
AirLab retweeted
Working on robotics and attending ICML 2026? (and also overwhelmed by the number of sessions like me?)
I made a website to help me navigate my first ICML with an emphasis on robotics papers!
Link: parvkpr.github.io/robotics-aโฆ
Sharing it here so people can use it. Check it out and let me know if you want me to add your paper/have any feedback!
Oh and also come checkout our poster on safety guardrails for navigation VLAs!
Link: constrained-robot-fms.githubโฆ
Weโll be presenting how constrained decoding can enforce safety requirments for robot navigation models at runtime, without retraining.
๐ Hall A #400
๐๏ธ Tue, July 7
โฐ 10:30โ11:30 AM KST
#icml #robotics #physicalai โ at Seoul, Republic of Korea
AirLab retweeted
What can different navigation domains (e.g. offroad, AV, humanoid, aerial, quadruped, digital) learn from each other?
๐ Excited to announce the Workshop on โBridging Perspectives in Navigationโ @ IROS 2026 (Pittsburgh, 9/27/2026).
Submit your paper by August 21! ๐
Proud to see Co-Me accepted to CVPR 2026 ๐
Now supporting MapAnything 1.1, Depth Anything 3, and Pi3 - and 2ร faster than the original, up to 21.5ร speedup on long VGGT sequences. Congrats to the team!
AirLab retweeted
Iโm excited to share that I successfully defended my Ph.D. thesis on Specification-Driven Planning for Safe Autonomy!
Iโm deeply grateful to my committee Sebastian Scherer, Changliu Liu, Karen Leung and Eunsuk Kang for their time and guidance throughout this journey. More in ๐งต
AirLab retweeted
Meet KinDER โ a stress test for robot physical reasoning. All 13 methods failed ๐
๐ 25 environments
โพ๏ธ Infinite tasks
๐๏ธ Gymnasium API
โ๏ธ Over 20 parameterized skills
๐ชง Human demonstrations
๐ 13 baselines (planning and learning)
From @Princeton @CMU_Robotics @ICatGT @CambridgeMLG @nvidia @MIT_CSAIL
๐งต 1/n
AirLab retweeted
I'm excited to share that RAVEN was accepted to ICRA 2026!
Paper: arxiv.org/abs/2509.23563
Website: raven-semantic.github.io
Collaboration with @OmarAlama, Dmytro Kurdydyk, John Keller, @Nik__V__ , Wenshan Wang, @ybisk , @smash0190
See you in Vienna!
AirLab retweeted
#IROS2026 will convene in Pittsburgh from Sept 27 โ Oct 1!
As one of the largest & most dynamic robotics conferences, IROS brings together world-leading researchers, educators, govt. leaders, startups, industry innovators, practitioners, & investors๐
2026.ieee-iros.org/
AirLab retweeted
Fast & light 2D and 3D zero-shot open-vocabulary semantic segmentation is here ๐๐ชถ!!
Meet RADSeg:
- 6-30% mIoU improvement while being 3.95x faster and using 2.5x fewer parameters.
- Outperforms combinations of huge vision models (850-1350M) with just 105M !
๐กThe key is building on the agglomerative model, RADIO, and improving spatial consistency.
AirLab retweeted
We are honored to share that Super Odometry is now published in @ScienceRobotics and featured as a highlight article! ๐ This work rethinks the SLAM paradigm: true resilience should not rely solely on external perceptionโit should begin from within.
science.org/stoken/author-toโฆ #SLAM
AirLab retweeted
Human peripheral vision reduces detail in out-of-focus areas. This โannoyingโ feature saves massive computation while preserving spatial cues. And for the most human-like artifact we buildโROBOTโthat efficiency matters.
Checkout our recent work:๐๐ co-me-tokens.github.io
