English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Learning goal conditioned control in the real world is a challenging openproblem in robotics. Reinforcement learning systems have the potential to learnautonomously via trial-and-error, but in practice the costs of manual rewarddesign, ensuring safe exploration, and hyperparameter tuning are often enoughto preclude real world deployment. Imitation learning approaches, on the otherhand, offer a simple way to learn control in the real world, but typicallyrequire costly curated demonstration data and lack a mechanism for continuousimprovement. Recently, iterative imitation techniques have been shown to learngoal directed control from undirected demonstration data, and improvecontinuously via self-supervised goal reaching, but results thus far have beenlimited to simulated environments. In this work, we present evidence thatiterative imitation learning can scale to goal-directed behavior on a realrobot in a dynamic setting: high speed, precision table tennis (e.g. "land theball on this particular target"). We find that this approach offers astraightforward way to do continuous on-robot learning, without complexitiessuch as reward design or sim-to-real transfer. It is also scalable -- sampleefficient enough to train on a physical robot in just a few hours. In realworld evaluations, we find that the resulting policy can perform on par orbetter than amateur humans (with players sampled randomly from a robotics lab)at the task of returning the ball to specific targets on the table. Finally, weanalyze the effect of an initial undirected bootstrap dataset size onperformance, finding that a modest amount of unstructured demonstration dataprovided up-front drastically speeds up the convergence of a general purposegoal-reaching policy. See https://sites.google.com/view/goals-eye for videos.

GoalsEye: Learning High Speed Precision Table Tennis on a Physical Robot