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 complex manipulation tasks in realistic, obstructed environments isa challenging problem due to hard exploration in the presence of obstacles andhigh-dimensional visual observations. Prior work tackles the explorationproblem by integrating motion planning and reinforcement learning. However, themotion planner augmented policy requires access to state information, which isoften not available in the real-world settings. To this end, we propose todistill a state-based motion planner augmented policy to a visual controlpolicy via (1) visual behavioral cloning to remove the motion plannerdependency along with its jittery motion, and (2) vision-based reinforcementlearning with the guidance of the smoothed trajectories from the behavioralcloning agent. We evaluate our method on three manipulation tasks in obstructedenvironments and compare it against various reinforcement learning andimitation learning baselines. The results demonstrate that our framework ishighly sample-efficient and outperforms the state-of-the-art algorithms.Moreover, coupled with domain randomization, our policy is capable of zero-shottransfer to unseen environment settings with distractors. Code and videos areavailable at https://clvrai.com/mopa-pd

Distilling Motion Planner Augmented Policies into Visual Control Policies for Robot Manipulation