Knowledge Extraction from Unknown Environment by Artificial-life Approach
Ryoji Sawa, Yuji Makita and Masafumi Hagiwara
Department of Electrial Engineering Faculty of Science and Technology, Keio University, 3-14-1 Hiyoshi, Kohoku- ku, Yokohama, 223-8522, Japan
Received:August 31, 1998Accepted:January 10, 1999Published:August 20, 1999
Keywords:Artificial life, Evolution, Knowledge extraction, Learning, Autonomous mobile robot, Emergence
Studies that target excellent information processing systems simulating life systems include the artificial-life (AL) incorporating emergence phenomenon - individual elements interacting based on lower level rules to generate higher level complex phenomena. Living things conserve species by replicating genes. We propose knowledge extraction promoting emergence by an AL approach incorporating living conservation of species and gene evolution. The proposed system consists of an AL environment and a knowledge extraction network. In the AL environment, individual elements interact and obtained data is input to the knowledge extraction network to present knowledge as a form of rules. Sets of rules are regarded as genes individual elements bequeath and new elements inherit these genes. We deal with a route-finding problem that, in simulation, sets a difficult situation whose goal is unknown, considering the actual world. Rules on route maps are extracted and extracted sets of rules are regarded as genes whose evolution is simulated through gene combination. We verified that the system finds routes effectively using evolved genes in a map made complex by combining maps.
Cite this article as:R. Sawa, Y. Makita, and M. Hagiwara, “Knowledge Extraction from Unknown Environment by Artificial-life Approach,” J. Adv. Comput. Intell. Intell. Inform., Vol.3 No.4, pp. 215-222, 1999.Data files: