Host: The Japan Society of Mechanical Engineers
Name : [in Japanese]
Date : June 01, 2022 - June 04, 2022
We propose a method that makes a mobile robot search for a target object using Bayesian estimation and natural language processing. When searching for an object, the robot must estimate where a person puts it. Otherwise, it takes too much time due to fruitless search. A previous research has proposed to implement human thinking processes on search to robots for efficient search. However, it is necessary to register the relation between each target object and the furniture where the object is put. Our method uses natural language processing to implement human thinking processes on search. The robot creates a map of the environment and divides the map to create an areas. Each area is evaluated by Bayesian inference and natural language processing. This allows robots to find objects efficiently. We have verified that this method works in a simulated environment built with Gazebo.
