Towards Safe and Reliable Robot Task Planning

Snehasis Banerjee

Abstract:

A long-standing goal of AI is to enable robots to
plan in dynamic and uncertain environments by
managing task failure intelligently. Reliability of a
robot task planner has become essential to ensure
operational safety. Reliability is generally determined by the probability of a task to circumvent
failures, while safety is related to the consequences
of the failures. In this paper, methods and an architecture for handling reliability and safety in robotic
task planning is presented. The paper takes help of
a telepresence navigation scenario to describe some
candidate solutions such as Task Reliability Graph
and weighted reliability goals.