A Research On Spatial Cognition And Representation Of Environment Based On Grey Qualitative Method

The cognition for environment is the basic precondition of mobile robot to mapping、locating、planning and moving in an unknown environment. The mobile robot not only should perceive the environment and extract the information, but also should have the capacity to understand、express and reason the environmental knowledge. It is the development tendency to cognize environment by simulating how human do. In consequence, it becomes the hot topics in the area of mobile robot about how to elimilate the uncertainty in perception, simulate the processing of human to spatial knowledge intelligently, guide the robot to cognize and express the information of an unknown、complicate and dynamic environment by an adaptive approach.Cognition is the process of operating on knowledge expression, i.e. generate, transform and delete. The development of cognition is the process of updating knowledge expression. The agent codes and simplifies the enviromental information as an expression which is called cognitive map through psychological function. In order to simulate human cognizing the environment, robot must understand the hierarchys of the environment, and express environmental knowledge by multi-resolution spatial expression pattern based on perceptual constancy. There are two topics shoud be solved:(1) How to express the different hierarchys of the environment? (2) How to understand and reason these expressions. In these topics, there are feature extraction、topological connetcion and spatial reasoning should be executed; the environmental information should be recognized、classified and conceptualized. To deal with the two issues based on collecting and collating relevant literatures and documents in recent years, this thesis focus on representation cognition and storage of environmental spacial knowledge based on grey qualitative method. This system could express the knowledge on different levels and multiple precisions by simulating the qualitative analysis ability of human efficiently to lay the foundation for the studying、reasoning and planning.The main contributions are as follows:(1) The concept and research target of cognitive map is confirmed by reviewing the previous research. As knowledge representation is the purpose, the necessity of solving uncertain environmental information qualitatively and the characters that a model of environmental representation should have has been analyzed. By combining the cognitive map with robot navigation, the theory system, and the mathematics foundation is discussed thoroughly based on the category, history and research subjects of cognitive map. The basic approaches and the key technical points to represent the knowledge in cognitive map are briefly reviewed. The development history of cognitive map is summarized. The developing trend is discussed by pointing the problem needing further research, and the significance of this paper has been elaborated.(2) The cognitive representation model based on grey qualitative simulation theory is proposed by simulating how people recognize an unknown environment. In the research of environmental cognition, the model proposed has the advantage of processing the incomplete knowledge in complex system through grey system theory. To decrease the complexity of the environmental model, the data bottom-up is processed on four levels:feature layer, adaptive layer, topological layer and cognitive layer. In order to plan path with prior knowledge, the up-bottom knowledge is applied on three levels:ontology, concept and object. To reduce the space complexity, the free space of local environment is decomposed into a group of convex polygons qualitatively by imitating human intelligence. The basic concepts of grey qualitative fundamental element, grey qualitative characteristic value, grey qualitative metric space and the grey qualitative relationship are defined.(3) The theoretical fundamental, implementation and the calculate model is researched. The influence of resolution ratio to grey correlation degree is discussed thoroughly by analyzing the computational method of grey correlation. The grey different information space of the environmental knowledge is established by taking advantage of grey correlation degree analysis, which can transform the relation of complex factors to the comparison between geometrical shapes of curves. The grey correlation degree of points and line-segment with each other are defined. The self-information and the common information in the point data can be abstracted through the grey correlation degree. As the representation of the feature layer, line-segments are detected by self-organizing mapping with main direction extracting based on grey correlation degree.(4) Based on research of the cognitive characters of human, the characters, mathematics fundamental, calculate model and significance of grey hazy set is discussed thoroughly. An environmental cognitive process based on the dynamic evolution of the grey hazy set is proposed. Based on the dynamic evolution of the grey hazy set, a cognitive algorithm is proposed for the knowledge expression of uncertain information, by simulating the gradually cognitive process of human. The algorithm produces different cognitive stage expression through dynamic evolution of the grey hazy set to eliminate the uncertainty of the information step by step. The example in corridor environment is provided to verify the effectiveness of the proposed algorithm in simulating human’s intelligence of environmental information expression and reasoning.(5) Aiming at the information processing on the adaptive layer of the cognitive representation model based on grey qualitative simulation theory, the mutual information is researched. An entropy based approach for processing environmental information is proposed based on information theory together with grey theory to eliminate the uncertainty in the information acquired by abstract the redundant information. The approach proposed simulates the process of human cognizing the environment to enhance the study of mobile robot from perception level to cognition level. In the process of the environmental cognition, the selected feature points are optimum in the sense that those features with existing information would yield the most entropy reduction. The approach proposed has the advantage of selecting a group of feature points to decrease the computation and storage cost in real time implementations.