Policy-Enhanced Fallback Nodes in Behavior Trees
Confirmation: I have read and agree with the workshop's policy on behalf of myself and my co-authors.
Keywords: Policy-Enhanced Fallback Nodes
Abstract: Robots operating in highly-dynamic environments
are burdened with the task of continuously monitoring and flexibly
adapting to their changing surroundings and internal state.
Fallback nodes in Behavior Trees offer a basic means to achieve
this adaptability, by determining a total static ordering over
contingency plans for achieving a goal. Attempting to encode a
more nuanced contextual ordering leads to significantly larger
Behavior Trees, sacrificing their human readability and the ease
of their elaboration. Towards maintaining these two desiderata,
we propose the enhancement of fallback nodes with symbolic
policies that transparently determine, at execution time, the
order in which contingency plans are to be considered according
to the current context. Based on our undertaken user study, we
offer evidence for the benefit of employing such policy-based
Behavior Trees over alternative versions of Behavior Trees.
Submission Number: 7
Loading