Spatial habituating self organizing map

Abstract

This paper presents the development of Spatial Habituating Self Organizing Map (SHSOM) network. This project is inspired by the challenges in underwater wall/pipe or cable inspection application using inspection robot. When exposed to the underwater natural elements, robot’s sensor readings are varied over space and time. Hence, the AUV need to be able to continuously adapt to its environment while performing inspection. For this reason, a new inspection system based on spatial Habituating Self Organizing Map (SHSOM) network is proposed. SHSOM allows the robot to continuously learn and adapt to new changes in its environment by using habituation principle which considers spatial information. WEBOT simulator is used to simulate an inspection scenario involving a mobile robot a changing environment. Simulation results show that the robot successfully learn and detect novel events during inspection.