The Methodology of State Space Construction with Self-organizing Incremental Neural Network on Subsumption Architecture
Fumiaki Saıtoh, Osamu Hasegawa · 한국지능시스템학회 국제학술대회 발표논문집 · 2009
The conventional Subsumption Architecture (SA) is not assumed to adjust to a specific environment. Therefore, it is necessary that SA adjust to environment. However, the robustness of SA is declined when robot adjusts to the specific environment. In this study, we proposed hybrid model of SA and Neural Network. The proposed model was consists of a Dynamics-Based Self-organizing Incremental Neural Network (DBSOINN) and SA. This model quickly adjusts to two or more environments because it used dynamics-based information processing. The effectiveness of the proposed method was verified using a simulation experiment in a maze problem.