Towards Self-Autonomy Evaluation using Behavior Trees

Abstract

Adjustable autonomy is an attractive paradigm to deploy autonomous robots that require occasional interaction with humans or partner robots. In multi-robot applications, autonomy levels such as teleoperation and fully autonomous are extended to team levels, and depending on the task might have complex hierarchical relations.This paper presents a preliminary work on multi-robot coordination strategy based on evaluating the robot’s level of autonomy. With appropriate assumptions, choosing the level of autonomy can be interpreted as a cooperative planning problem. To this end, we propose to encode the robot’s task and motion plans as a Behavior Tree (BT) to monitor the execution and react to external disturbances. Our approach combines an informative path planning with BT synthesis to obtain a plan that allows a robot to explore and act depending on the uncertainty in an environment representation. We demonstrate how a robot can switch between the single and cooperative execution depending on the exploration’s outcome in a navigation among movable objects scenario.