ACTIVE LOCALIZATION OF ROBOT USING SPATIAL SEMANTICS OF OBJECTS

摘要

PURPOSE: An active self-location estimation method of a robot using spatial semantic information of objects is provided to reduce the calculation of a robot and power consumption. CONSTITUTION: An active self-location estimation method of a robot using spatial semantic information of objects is as follows. A finite number of virtual behaviors, a finite number of reference objects, and a finite number of spatial semantic information are specified, and the spatial semantic information defines a spatial context where at least one object is placed(S1). Stochastic reliability on the position of the robot is obtained(S2). The expectation of the entropy of virtual position reliability for all combinations of the spatial semantic information and objects is calculated every virtual behavior(S3). A virtual behavior, which corresponds to a minimum value of the calculated expectations, is determined(S4).