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 Tools annual

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 Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

We focus on the task of goal-oriented grasping, in which a robot is supposedto grasp a pre-assigned goal object in clutter and needs some pre-grasp actionssuch as pushes to enable stable grasps. However, in this task, the robot getspositive rewards from environment only when successfully grasping the goalobject. Besides, joint pushing and grasping elongates the action sequence,compounding the problem of reward delay. Thus, sample inefficiency remains amain challenge in this task. In this paper, a goal-conditioned hierarchicalreinforcement learning formulation with high sample efficiency is proposed tolearn a push-grasping policy for grasping a specific object in clutter. In ourwork, sample efficiency is improved by two means. First, we use agoal-conditioned mechanism by goal relabeling to enrich the replay buffer.Second, the pushing and grasping policies are respectively regarded as agenerator and a discriminator and the pushing policy is trained withsupervision of the grasping discriminator, thus densifying pushing rewards. Todeal with the problem of distribution mismatch caused by different trainingsettings of two policies, an alternating training stage is added to learnpushing and grasping in turn. A series of experiments carried out in simulationand real world indicate that our method can quickly learn effective pushing andgrasping policies and outperforms existing methods in task completion rate andgoal grasp success rate by less times of motion. Furthermore, we validate thatour system can also adapt to goal-agnostic conditions with better performance.Note that our system can be transferred to the real world without anyfine-tuning. Our code is available athttps://github.com/xukechun/Efficient_goal-oriented_push-grasping_synergy.

Efficient learning of goal-oriented push-grasping synergy in clutter