2A1-M02 形態と知能を有する進化ロボットに関する研究(進化・学習とロボティクス)

概要

論文の詳細を見る
This paper deals with a concept that effective robotic function emerges from not only "intelligence" but also the balance of "intelligence" and "morphology" with physical interaction in the environment through the task. In this paper, the "morphology" of a robot is represented by the arrangement of two visual sensors and the "intelligence" is described as a decision tree of sensor inputs and motor outputs. Both of "morphology" and "intelligence" are automatically generated and evolved by genetic programming (GP) for a task of maintaining a certain distance between the robot and an object. The result shows that noise existing during a translation process from a phenotype to genotype doubles the incidence of symmetry in "Intelligence", i.e., a state-action coordination and an evolutionary robotic system with symmetry in "Intelligences" accomplishes a given task ever real world, because it has high robustness.
一般社団法人日本機械学会の論文
2007-05-11