Tim Missal
Tim Missal is a graduate researcher at the Technical University of Darmstadt, affiliated with the Intelligent Autonomous Systems (IAS) Lab led by Professor Jan Peters. He is best known as a co-first author of RopeDreamer, a deep learning framework that predicts the dynamics of deformable linear objects such as ropes and cables using a kinematic recurrent state space model. The work, submitted to arXiv in April 2026, demonstrates a 40.52% reduction in open-loop prediction error at 50-step horizons and a 31% improvement in inference speed over the prior state of the art, with potential applications in robotic manipulation of flexible objects.
“RopeDreamer proposes a new architecture that predicts flexible object behavior 40% more accurately over long horizons while running 31% faster than the current best approach.”
Source→“RopeDreamer's best model, by contrast, accumulates only 19.05mm of error at t=50. That's a 40.52% reduction in prediction error at the 50-step horizon — and the gap widens over time, not narrows.”
Source→“The simulation is implemented in MuJoCo 3.3.7, where the DLO is modeled as a chain of 70 capsules with a length of 10mm and a thickness of 10mm, connected by ball joints.”
Source→AI-extracted from podcast / newsletter / paper summaries. May contain errors.