FUNCTIONAL MAPPING FOR HUMAN-ROBOT COLLABORATIVE EXPLORATION

摘要

Our problem is one of a human-robot team exploring arnpreviously unknown disaster scenario together. The teamrnis building up situation awareness, gathering informationrnabout the prescence and structure of specific objects of interestrnlike victims or threats. For a robot working with arnhuman team, there are several challenges. From the viewpointrnof task-work, there is time-pressure: The explorationrnneeds to be done efficiently, and effectively. From the viewpointrnof team-work, the robot needs to perform its tasksrntogether with the human users such that it is apparent tornthe users why the robot is doing what it is doing. Withoutrnthat, human users might fail to trust the robot, whichrncan negatively impact overall team performance. In thisrnpaper, we present an approach to the field of semantic mapping,rnas a subset of robotic mapping; aiming to address thernproblems in both efficiency (task), and apparency (team).rnThe approach models the environment from a geometricalfunctionalrnviewpoint, establishing where the robot needsrnto be, to be in an optimal position to gather particular informationrnrelative to a 3D-landmark in the environment.rnThe approach combines top-down logical and probabilisticrninferences about 3D-structure and robot morphology,rnwith bottom-up quantitative maps. The inferences result inrnvantage positions for information gathering which are optimalrnin a quantitative sense (effectivity), and which mimicrnhuman spatial understanding (apparency). A quantitativernevaluation shows that functional mapping leads to significantlyrnbetter vantage points than a naive approach.