Simulated decontamination experiments by mobile manipulator with visual recognition - Autonomous behavior driven by environment programming

Abstract

Since Tohoku Earthquake off the coast of Pacific Ocean at Fukushima Prefecture and subsequent Tsunami devastated nuclear power station, No.1, in Fukushima, radioactive substances have been scattered off wide area of Fukushima. Then decontamination works have started, but they have not been progressing rapidly because there are not sufficient workers and they have risks that they should be exposed to radiation, which restrict working duration. So this research aims at developing a robot that moves and collects polluted soil by itself. This machine, controlled by visual information, recognizes and collects simulated contaminated soils automatically, where Model-based Matching Method and Genetic Algorithm are used to recognize the contaminated-soil-model. In this paper, recognition methods, structure of mobile manipulator with hand-eye cameras and results of simulated decontamination experiments are reported, in which a concept of autonomous behavior driven by ``Environment Programming,'' meaning that environment gives programmed commands to the autonomous mobile manipulator.