IN PROGRESS · TARGET: SEP 2026
Wiggle and Throw!
Real-to-sim sensor alignment through demonstration contrasting
Email me to discuss →Arthur Jakobsson
Robotics Researcher @ CMU
Birds, photography, and computer science.
Hi! I'm a full-time masters student at Carnegie Mellon University working in the Momentum Lab with Professor Jeff Ichnowski. I'm applying to PhD programs in computer vision and robotics for Fall 2026, and I'm specifically excited about machine learning-enabled soft and articulated object manipulation, high-speed manipulation, and creative sensing. I'm also currently a research intern at DENSO, working on out-of-distribution 4D scene representations for autonomous vehicles (details under NDA).
My most recent work is Wiggle and Go, a system that enables a robot to swing ropes at 3D targets zero-shot: by observing a few seconds of rope motion, predicting physical parameters, and conditioning a goal-directed policy on them. In my undergrad, I was part of the SBPL Lab, advised by Professor Maxim Likhachev where I worked on machine learning for multi-agent path-finding.
From 2022 to 2025 I served as a TA, lead TA, and then course instructor, for 15-122 Principles of Imperative Computation. I built and ran the bootcamp series now used across the course — over 3,000 cumulative student attendances. Teaching matters to me and I hope to get to keep teaching as I continue my academic career!
Outside of research I love birdwatching, photography, biking, and badminton.
Preprint · 2026 · First-author
Wiggle and Go.
A single mistake in a dynamic throw can be unrecoverable. We use a brief wiggle to identify rope dynamics, then perform goal-conditioned manipulation in one shot.
In progress
What's brewing
Two upcoming first-author projects with the Momentum Lab and BIG Lab at CMU. Reach out if you'd like to chat about either — get in touch →
RESEARCH THREAD
Active sensing for articulated objects
Robots in the loop for articulation discovery
A series on robots that close the loop on perception how to choose demonstrations and sensor sets that disambiguate joint structure rather than treating articulation as a passive perception problem.
Email me to discuss →
Updates
News
- First-author paper RareOcc accepted to the SDAD workshop at ECCV 2026 — controllable 4D occupancy for long-tail driving scenes. Keep an eye out for a full paper follow-up.
- Second-author preprint posted: Intercepting the Future — AHEAD, a latent-space predictive world model for dynamic VLA manipulation.
- Excited to be starting my internship at DENSO Innovation Lab this summer! I will be working on research in AV vision and perception with simulations.
- First-author preprint posted: Wiggle and Go! — zero-shot dynamic rope manipulation with the Momentum Lab.
- Two co-authored papers accepted to ICRA 2026: ExpReS-VLA and Adversarial Game-Theoretic Grasp Synthesis.
- Started full-time masters at Carnegie Mellon, focused on computer vision and robotics.
- Co-instructor with Mihir Khare for the CMU summer session of 15-122 Principles of Imperative Computation (120 students).
- ICRA 2025 paper accepted: Work Smarter Not Harder with the SBPL Lab.
- Wrapped up four years as TA for 15-122 (2022–2025).
- ICAPS 2024 paper out: Improving Learnt Local MAPF Policies with Heuristic Search.
- Received the "Coolest Graphs" award in 36-402 Statistical Computing: Manrattan - A Look into NYC's Rats .