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 from human demonstrations can facilitate automation but is riskybecause the execution of the learned policy might lead to collisions and otherfailures. Adding explicit constraints to avoid unsafe states is generally notpossible when the state representations are complex. Furthermore, enforcingthese constraints during execution of the learned policy can be challenging inenvironments where dynamics are difficult to model such as push mechanics ingrasping. In this paper, we propose Derivative-Free Recovery (DFR), a two-phasemethod for generating robust policies from demonstrations in roboticmanipulation tasks where the system comes to rest at each time step. In thefirst phase, we use support estimation of supervisor demonstrations and treatthe support as implicit constraints on states. We also propose a time-varyingmodification for sequential tasks. In the second phase, we use this supportestimate to derive a switching policy that employs the learned policy in theinterior of the support and switches to a recovery policy to steer the robotaway from the boundary of the support if it drifts too close. We presentadditional conditions, which linearly bound the difference in state at eachtime step by the magnitude of control, allowing us to prove that the robot willnot violate the constraints using the recovery policy. A simulated pushing taskin MuJoCo suggests that DFR can reduce collisions by 83\%. On a physical linetracking task using a da Vinci Surgical Robot and a moving Stewart platform,DFR reduced collisions by 84\%.

Constraint Estimation and Derivative-Free Recovery for Robot Learning from Demonstrations