Matches in SemOpenAlex for { <https://semopenalex.org/work/W4362722144> ?p ?o ?g. }
Showing items 1 to 90 of
90
with 100 items per page.
- W4362722144 abstract "Observing a human demonstrator manipulate objects provides a rich, scalable and inexpensive source of data for learning robotic policies. However, transferring skills from human videos to a robotic manipulator poses several challenges, not least a difference in action and observation spaces. In this work, we use unlabeled videos of humans solving a wide range of manipulation tasks to learn a task-agnostic reward function for robotic manipulation policies. Thanks to the diversity of this training data, the learned reward function sufficiently generalizes to image observations from a previously unseen robot embodiment and environment to provide a meaningful prior for directed exploration in reinforcement learning. We propose two methods for scoring states relative to a goal image: through direct temporal regression, and through distances in an embedding space obtained with time-contrastive learning. By conditioning the function on a goal image, we are able to reuse one model across a variety of tasks. Unlike prior work on leveraging human videos to teach robots, our method, Human Offline Learned Distances (HOLD) requires neither a priori data from the robot environment, nor a set of task-specific human demonstrations, nor a predefined notion of correspondence across morphologies, yet it is able to accelerate training of several manipulation tasks on a simulated robot arm compared to using only a sparse reward obtained from task completion." @default.
- W4362722144 created "2023-04-09" @default.
- W4362722144 creator A5008880429 @default.
- W4362722144 creator A5022706238 @default.
- W4362722144 creator A5045217258 @default.
- W4362722144 creator A5062817741 @default.
- W4362722144 creator A5076374279 @default.
- W4362722144 date "2023-05-29" @default.
- W4362722144 modified "2023-10-14" @default.
- W4362722144 title "Learning Reward Functions for Robotic Manipulation by Observing Humans" @default.
- W4362722144 cites W2051228319 @default.
- W4362722144 cites W2194775991 @default.
- W4362722144 cites W2605102758 @default.
- W4362722144 cites W2625366777 @default.
- W4362722144 cites W2769112066 @default.
- W4362722144 cites W2962736495 @default.
- W4362722144 cites W2962760690 @default.
- W4362722144 cites W2963703448 @default.
- W4362722144 cites W3009295642 @default.
- W4362722144 cites W3035198432 @default.
- W4362722144 cites W3122520957 @default.
- W4362722144 cites W3189615635 @default.
- W4362722144 cites W3205786327 @default.
- W4362722144 cites W3207832698 @default.
- W4362722144 cites W4214612132 @default.
- W4362722144 doi "https://doi.org/10.1109/icra48891.2023.10161178" @default.
- W4362722144 hasPublicationYear "2023" @default.
- W4362722144 type Work @default.
- W4362722144 citedByCount "0" @default.
- W4362722144 crossrefType "proceedings-article" @default.
- W4362722144 hasAuthorship W4362722144A5008880429 @default.
- W4362722144 hasAuthorship W4362722144A5022706238 @default.
- W4362722144 hasAuthorship W4362722144A5045217258 @default.
- W4362722144 hasAuthorship W4362722144A5062817741 @default.
- W4362722144 hasAuthorship W4362722144A5076374279 @default.
- W4362722144 hasBestOaLocation W43627221442 @default.
- W4362722144 hasConcept C107457646 @default.
- W4362722144 hasConcept C119857082 @default.
- W4362722144 hasConcept C14036430 @default.
- W4362722144 hasConcept C154945302 @default.
- W4362722144 hasConcept C162324750 @default.
- W4362722144 hasConcept C177264268 @default.
- W4362722144 hasConcept C187736073 @default.
- W4362722144 hasConcept C199360897 @default.
- W4362722144 hasConcept C2780451532 @default.
- W4362722144 hasConcept C41008148 @default.
- W4362722144 hasConcept C41608201 @default.
- W4362722144 hasConcept C48044578 @default.
- W4362722144 hasConcept C77088390 @default.
- W4362722144 hasConcept C78458016 @default.
- W4362722144 hasConcept C86803240 @default.
- W4362722144 hasConcept C90509273 @default.
- W4362722144 hasConcept C97541855 @default.
- W4362722144 hasConceptScore W4362722144C107457646 @default.
- W4362722144 hasConceptScore W4362722144C119857082 @default.
- W4362722144 hasConceptScore W4362722144C14036430 @default.
- W4362722144 hasConceptScore W4362722144C154945302 @default.
- W4362722144 hasConceptScore W4362722144C162324750 @default.
- W4362722144 hasConceptScore W4362722144C177264268 @default.
- W4362722144 hasConceptScore W4362722144C187736073 @default.
- W4362722144 hasConceptScore W4362722144C199360897 @default.
- W4362722144 hasConceptScore W4362722144C2780451532 @default.
- W4362722144 hasConceptScore W4362722144C41008148 @default.
- W4362722144 hasConceptScore W4362722144C41608201 @default.
- W4362722144 hasConceptScore W4362722144C48044578 @default.
- W4362722144 hasConceptScore W4362722144C77088390 @default.
- W4362722144 hasConceptScore W4362722144C78458016 @default.
- W4362722144 hasConceptScore W4362722144C86803240 @default.
- W4362722144 hasConceptScore W4362722144C90509273 @default.
- W4362722144 hasConceptScore W4362722144C97541855 @default.
- W4362722144 hasLocation W43627221441 @default.
- W4362722144 hasLocation W43627221442 @default.
- W4362722144 hasLocation W43627221443 @default.
- W4362722144 hasLocation W43627221444 @default.
- W4362722144 hasLocation W43627221445 @default.
- W4362722144 hasOpenAccess W4362722144 @default.
- W4362722144 hasPrimaryLocation W43627221441 @default.
- W4362722144 hasRelatedWork W2031695474 @default.
- W4362722144 hasRelatedWork W2138720691 @default.
- W4362722144 hasRelatedWork W2389214306 @default.
- W4362722144 hasRelatedWork W2586732548 @default.
- W4362722144 hasRelatedWork W3049728571 @default.
- W4362722144 hasRelatedWork W3213722473 @default.
- W4362722144 hasRelatedWork W4306904969 @default.
- W4362722144 hasRelatedWork W4362501864 @default.
- W4362722144 hasRelatedWork W4380318855 @default.
- W4362722144 hasRelatedWork W3111219495 @default.
- W4362722144 isParatext "false" @default.
- W4362722144 isRetracted "false" @default.
- W4362722144 workType "article" @default.