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

Sim-to-real transfer is a powerful paradigm for robotic reinforcementlearning. The ability to train policies in simulation enables safe explorationand large-scale data collection quickly at low cost. However, prior works insim-to-real transfer of robotic policies typically do not involve anyhuman-robot interaction because accurately simulating human behavior is an openproblem. In this work, our goal is to leverage the power of simulation to trainrobotic policies that are proficient at interacting with humans upondeployment. But there is a chicken and egg problem -- how to gather examples ofa human interacting with a physical robot so as to model human behavior insimulation without already having a robot that is able to interact with ahuman? Our proposed method, Iterative-Sim-to-Real (i-S2R), attempts to addressthis. i-S2R bootstraps from a simple model of human behavior and alternatesbetween training in simulation and deploying in the real world. In eachiteration, both the human behavior model and the policy are refined. For alltraining we apply a new evolutionary search algorithm called Blackbox GradientSensing (BGS). We evaluate our method on a real world robotic table tennissetting, where the objective for the robot is to play cooperatively with ahuman player for as long as possible. Table tennis is a high-speed, dynamictask that requires the two players to react quickly to each other's moves,making for a challenging test bed for research on human-robot interaction. Wepresent results on an industrial robotic arm that is able to cooperatively playtable tennis with human players, achieving rallies of 22 successive hits onaverage and 150 at best. Further, for 80% of players, rally lengths are 70% to175% longer compared to the sim-to-real plus fine-tuning (S2R+FT) baseline. Forvideos of our system in action, please see https://sites.google.com/view/is2r.

i-Sim2Real: Reinforcement Learning of Robotic Policies in Tight Human-Robot Interaction Loops