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

Dynamic tasks like table tennis are relatively easy to learn for humans butpose significant challenges to robots. Such tasks require accurate control offast movements and precise timing in the presence of imprecise state estimationof the flying ball and the robot. Reinforcement Learning (RL) has shown promisein learning of complex control tasks from data. However, applying step-based RLto dynamic tasks on real systems is safety-critical as RL requires exploringand failing safely for millions of time steps in high-speed regimes. In thispaper, we demonstrate that safe learning of table tennis using model-freeReinforcement Learning can be achieved by using robot arms driven by pneumaticartificial muscles (PAMs). Softness and back-drivability properties of PAMsprevent the system from leaving the safe region of its state space. In thismanner, RL empowers the robot to return and smash real balls with 5 m\s and12m\s on average to a desired landing point. Our setup allows the agent tolearn this safety-critical task (i) without safety constraints in thealgorithm, (ii) while maximizing the speed of returned balls directly in thereward function (iii) using a stochastic policy that acts directly on thelow-level controls of the real system and (iv) trains for thousands of trials(v) from scratch without any prior knowledge. Additionally, we present HYSR, apractical hybrid sim and real training that avoids playing real balls duringtraining by randomly replaying recorded ball trajectories in simulation andapplying actions to the real robot. This work is the first to (a) fail-safelearn of a safety-critical dynamic task using anthropomorphic robot arms, (b)learn a precision-demanding problem with a PAM-driven system despite thecontrol challenges and (c) train robots to play table tennis without realballs. Videos and datasets are available at muscularTT.embodied.ml.

Learning to Play Table Tennis From Scratch using Muscular Robots