In the present work, a computer model of an autonomous agent that forms its own knowledge base is constructed and studied. We consider a rather simple knowledge base of an agent, which forms and uses a cognitive map of the external world. This map is the agent knowledge base. The external world of an agent is a two-dimensional cellular environment. In this environment, the agent can eat and move. The agent prefers certain cells for eating. Namely, such cells form the nutrient spots of the world. The model was investigated by means of computer simulation. We analyze the processes of formation of a cognitive map in the agent knowledge base that characterizes the nutrient spots of the world.