[Abstract]Faced with complex and unstructured operating environments, knowledge representation methods play an increasingly important role in robot autonomous operations. Knowledge representation focuses on the representation of knowledge symbols and how to automatically realize knowledge processing through reasoning procedures. Based on this, the robot knowledge representation framework and the latest application progress based on ontology representation and reasoning are introduced. From the two aspects of deterministic knowledge and uncertain knowledge, the technical background, implementation methods and research status of knowledge representation and reasoning are introduced, and the future research direction of robot knowledge is prospected.
[Keywords] Knowledge representation; knowledge reasoning; ontology; knowledge-based robot
0 Preface
Benefiting from the rapid development of artificial intelligence and robotics technology, the perception capabilities of robots have been significantly improved, and robots can complete some basic object recognition, navigation and operation tasks. However, for operations in complex and changeable environments, such as home service robots for table setting and human-machine collaborative robot assembly services, robots lack adaptability in the process of performing tasks. Facing human-scale, incomplete or fuzzy task problems, it is difficult for robots to adjust their action strategies in response to environmental changes, and it is difficult to achieve autonomous decomposition or planning of complex tasks to robot atomic actions. At the same time, in the process of interacting with humans, although service robots already have strong speech recognition and question and answer capabilities, the lack of semantic understanding capabilities makes it difficult for robots to map words in natural language to the operating environment, such as object position relationships, status, Functional attributes, etc. The lack of cognitive ability has become a key bottleneck restricting the development of robots. The specific manifestations are: robots lack awareness of their own operating capabilities, and their awareness of the environment mostly stays at the object detection level. Relatively simple information such as object classification and positioning is difficult to Meet the requirements for further understanding target items and operating target items to provide complex scene operations.
Artificial intelligence symbolism believes that the cognitive process is a symbol processing process in ontology, and the human thinking process can always be described with some symbol. The concept of ontology originates from the field of philosophy, where it is defined as a systematic description of objective things in the world. The field of artificial intelligence mainly uses the concept of ontology to express knowledge, that is, the basic elements in ontology (concepts and relationships between concepts) are used to describe the real world.