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

Multi-goal robot manipulation tasks with sparse rewards are difficult forreinforcement learning (RL) algorithms due to the inefficiency in collectingsuccessful experiences. Recent algorithms such as Hindsight Experience Replay(HER) expedite learning by taking advantage of failed trajectories andreplacing the desired goal with one of the achieved states so that any failedtrajectory can be utilized as a contribution to learning. However, HERuniformly chooses failed trajectories, without taking into account which onesmight be the most valuable for learning. In this paper, we address this problemand propose a novel approach Contact Energy Based Prioritization~(CEBP) toselect the samples from the replay buffer based on rich information due tocontact, leveraging the touch sensors in the gripper of the robot and objectdisplacement. Our prioritization scheme favors sampling of contact-richexperiences, which are arguably the ones providing the largest amount ofinformation. We evaluate our proposed approach on various sparse reward robotictasks and compare them with the state-of-the-art methods. We show that ourmethod surpasses or performs on par with those methods on robot manipulationtasks. Finally, we deploy the trained policy from our method to a real Frankarobot for a pick-and-place task. We observe that the robot can solve the tasksuccessfully. The videos and code are publicly available at:https://erdiphd.github.io/HER_force

Contact Energy Based Hindsight Experience Prioritization