Journal Article 人工生物の形態進化における獲得形態の評価 —タスク難易度が及ぼす獲得形態への影響—

岩館, 健司  ,  鈴木, 育男  ,  渡辺, 美知子  ,  山本, 雅人  ,  古川, 正志  ,  IWADATE, Kenji  ,  SUZUKI, Ikuo  ,  WATANABE, Michiko  ,  YAMAMOTO, Masahito  ,  FURUKAWA, Masashi

82 ( 6 )  , pp.595 - 601 , 2016-06 , 公益社団法人 精密工学会
This study considered the influence of task complexity on evolution of virtual creatures. Creatures consist of rigid primitives, joints, and neural controller and these parameters are evolved with real number coded genetic algorithm. Creatures are evaluated in a virtual environment that implemented physical forces for achieving various simple tasks such as walking, pursuit toward a target, and evasion from obstacles. Acquired shapes of creatures are classified into 4 types according to a feature value named "number of branches". Experimental results showed that the creature has a specific number of branches obtained a high fitness.

Number of accesses :  

Other information