A Review of Domain Knowledge Representation for Robot Task Planning

摘要

The review analyzes and defines the related key concepts firstly. Then the article reviews the knowledge representation (KR) formalism and the existing knowledge systems in robotics. The review divides KR in robotics into three categories, namely semantic networks, rules and logical knowledge representation. Then the existing knowledge systems are classified according to the three categories. Finally the article introduces three typical knowledge representation languages in the application of robot task planning and the corresponding task planners. It is still a disadvantage that the domain knowledge of the existing task planners is generated manually, so that the article introduces the method of machine learning to generate the domain knowledge automatically. Our contributions are to facilitate the selection of a knowledge system according to the requirements of robotic application and predict the development trend, hoping to inspire and promote the research in autonomous robotics.