Multi-knowledge for robot to identify environments

An important topic in robotics is how a mobile robot perceives its environment and determines its location within this environment. Typically, the environment is represented by a feature decision system and techniques such as machine learning or data mining are used for identification of the environment. However, conventional representations with a single body of knowledge encounters many problems when the environment is changed. In this paper, multi-knowledge is defined by means of mapping vector spaces and is used to tackle the problem of robot environment identification. It is shown that a robot with multi-knowledge is capable of identifying changes in environment. The multi-knowledge approach is based on the multi-reducts of the environment feature decision system. In order to find multi-reducts, an algorithm based on the rough set theory is proposed in this paper. The algorithm has been used to find the multi-reducts in the data sets from UCI machine learning repository. The experimental results show that the algorithm is efficient and that most data sets in the real word have multi-reducts. This paper shows that not only does multi-knowledge can be used in the example presented but that it has a wide range of application areas.