Samford University1, Computer Science, Birmingham, AL 35229 Samford2, Computer Science, Birmingham, AL 35229
Humans have used their ability to navigate environments since they came into existence. Though this skill is considered an integral part of human survival, the exact understanding of human navigation remained either unknown or disregarded until the early 1940s. Through psychological experiments using rodents, scientists began to form theories regarding spatial cognition and navigation. These same ideas, in conjunction with current technological understanding, have given rise to computer models which can supplement abstract theories. Our goal was to investigate the formation of a cognitive map using our own computer model based on previous research and computer simulation. To accomplish this task of using an artificial system to model human way-finding, we needed to accomplish much of what the human brain normally does for us. Camera images are preprocessed for landmark identification and learned via use of a neural network. Based on limited views of the environment, the system builds proximity relations between pairs of landmarks, which become the building blocks of the artificial cognitive map. Relationships between landmarks not seen in the same environmental view are deduced via the transitive property and used to navigate the environment. This model has been implemented in a robot which autonomously learns a two dimensional environment and has the ability to physically navigate it. These kinds of models hold the promise of greater understanding of human spatial cognition.