aiqbench analysis
This research advances robotic manipulation by training policies to actively resolve kinematic ambiguities through learned informative exploration rather than greedy search. By integrating Bayesian filtering with reinforcement learning, the method effectively handles objects where visual appearance misleads standard perception models. The approach demonstrates significant gains on benchmarks involving complex articulated structures like doors and drawers.
Learning to Explore Hidden Kinematics for Articulated Object Manipulation
arXiv:2609.36553v1 Announce Type: new Abstract: The kinematics of an articulated object is often ambiguous from vision alone. Interaction resolves the ambiguity, and active perception methods exploit this by searching for the single action that most sharpens a belief over the kinematic parameters at each step. Such greedy search cannot be extended over a horizon without forward models of the contact and inertial dynamics, which are themselves unknown. We instead amortize action selection into training. We maintain a belief distribution over joint type and parameters, initialized from a generative prior and updated by Bayesian filtering on the observed part motion. To condition the policy on this belief, we render it as a per-point articulation flow field, the motion that the current posterior predicts for every point on the object. Carrying the inductive bias of articulated motion, this representation generalizes better than a latent encoding of the belief or flow tracked from observation. We train the policy with reinforcement learning, rewarding the entropy that each interaction removes from the posterior, so that informative exploration becomes learned behavior rather than a search at every step. Our method outperforms previous approaches across door and drawer manipulation on the PartManip benchmark, and reaches 61.7% success on ArticuRiddle, a new dataset of objects whose appearance implies the wrong articulation, against 44.4% for the best previous method. Project Website: https://hiddenkinematics.github.io/
Abstract:The kinematics of an articulated object is often ambiguous from vision alone. Interaction resolves the ambiguity, and active perception methods exploit this by searching for the single action that most sharpens a belief over the kinematic parameters at each step. Such greedy search cannot be extended over a horizon without forward models of the contact and inertial dynamics, which are themselves unknown. We instead amortize action selection into training. We maintain a belief distribution over joint type and parameters, initialized from a generative prior and updated by Bayesian filtering on the observed part motion. To condition the policy on this belief, we render it as a per-point articulation flow field, the motion that the current posterior predicts for every point on the object. Carrying the inductive bias of articulated motion, this representation generalizes better than a latent encoding of the belief or flow tracked from observation. We train the policy with reinforcement learning, rewarding the entropy that each interaction removes from the posterior, so that informative exploration becomes learned behavior rather than a search at every step. Our method outperforms previous approaches across door and drawer manipulation on the PartManip benchmark, and reaches 61.7% success on ArticuRiddle, a new dataset of objects whose appearance implies the wrong articulation, against 44.4% for the best previous method. Project Website: this https URL
Robotics (cs.RO)
arXiv:2609.36553 [cs.RO]
(or arXiv:2609.36553v1 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2609.36553 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 313 words
Related coverage
Game Design Creativity evals
LessWrong · Oct 12, 2026
Try Once, Then Optimal: De-Redundified Procedure Memory for Cross-Episode Exploration Amortization
arXiv · Jul 28, 2026
Latent reasoning architectures would likely undermine CoT, our strongest oversight tool
AI Alignment Forum · Sep 23, 2026
Beyond Dexterity: Why Contact May Define the Next Era of Robotics
IEEE Spectrum · Jun 9, 2026