DexHOT: Learning Hand-Object Configuration Transfer for Dexterous Manipulation
DexHOT: Learning Hand-Object Configuration Transfer for Dexterous Manipulation
This research introduces DexHOT, a framework that formulates dexterous manipulation as hand-object configuration transfer. The method learns a reusable relation-conditioned transfer primitive from synthetic feasible configuration pairs without human reference motions and composes long-horizon behaviors through closed-loop tracking of configuration subgoals. On DexHOTBench with 120 objects, a single policy achieves 80.5% transfer success on seen objects and 69.4% on unseen objects; on a real dexterous hand, it achieves 61.5% success over 200 real-robot transfers and 90% task success across six demonstration-guided tasks.
The paper introduces DexHOT, a method for dexterous manipulation that learns hand-object configuration transfer from synthetic data without human reference motions, achieving 80.5% success on seen objects and 69.4% on unseen objects in simulation, with 61.5% success over 200 real-robot transfers and 90% task success across six demonstration-guided tasks.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
number of objects on DexHOTBench
120
View original evidence
as hand-object configuration transfer: moving between feasible configurations that jointly specify hand articulation and object pose relative to the wrist. DexHOT learns a reusable relation-conditioned transfer primitive from synthetic feasible configuration pairs, without human reference motions, and composes long-horizon behaviors through closed-loop tracking of configuration subgoals. On DexHOTBench with 120 objec
Open source S3number of demonstration-guided tasks
6 episodes
Reported trials
View original evidence
ts, a single policy achieves 80.5% transfer success on seen objects and 69.4% on unseen objects; on a real dexterous hand, it achieves 61.5% success over 200 real-robot transfers and 90% task success across six demonstration-guided tasks. These results show that explicit hand-object goals provide an expressive and composable interface for dexterous manipulation. More details and videos are available at https://dex-ho
Open source S4
- paper: not_reported
- code: not_reported
- weights: not_reported
- dataset: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
ts, a single policy achieves 80.5% transfer success on seen objects and 69.4% on unseen objects; on a real dexterous hand, it achieves 61.5% success over 200 real-robot transfers and 90% task success across six demonstration-guided tasks. These results show that explicit hand-object goals provide an expressive and composable interface for dexterous manipulation. More details and videos are available at https://dex-ho
Open source S4
as hand-object configuration transfer: moving between feasible configurations that jointly specify hand articulation and object pose relative to the wrist. DexHOT learns a reusable relation-conditioned transfer primitive from synthetic feasible configuration pairs, without human reference motions, and composes long-horizon behaviors through closed-loop tracking of configuration subgoals. On DexHOTBench with 120 objec
Open source S3
Source：arXiv Robotics — research abstracts · arxiv.org