AMOS; active perception of an autonomous system

Abstract

In the autonomous mobile systems project (AMOS), the FAW uses a mobile robot to study questions related to the deep integration of sub-symbolic and symbolic information processing. AMOS aims at methods for autonomously acquiring new concepts via induction from its interaction with its environment. This paper presents an architecture which integrates both symbolic planning as well as nonsymbolic reactive mechanisms, thus providing a basic autonomy, so that the robot can freely maneuver around without any detailed model of itself and its complex real-world environment . Substantial differences between expectation and observation are used as hints-generated via the robot's interaction with the environment-to situations which are of relevance for the robot. In particular, the concepts of plan breakdown and region of interest play a fundamental role. Autonomously, based on the robot's decision, images are taken and clustered without supervision into groups which are expected to correspond to semantically similar situations. These hypotheses shall be used in further work as a necessary pre-requisite in order to autonomously generate a new concept relating the recognition of such perception classes to appropriate actions.<>