A novel neurophysiological based navigation system

This paper introduces a novel neurophysiologically based mobile robot navigation system, which emulates the dynamics of a rodent's navigation and spatial awareness cells found in the hippocampus and entorhinal cortex. The model presented here replicates the functionality of these neurons in their hardware and software counterparts. By using data structures and computational logic that best utilizes currently available processing architectures, a cognitive map is created using a unique multimodal source model for place cell activation. Path planning is performed by using a combination of Euclidean distance path checking, goal memory, and the A∗ algorithm. Localization is accomplished using simple, low power sensors, such as a camera, ultrasonic sensors, motor encoders and a gyroscope. The place code data structures are initialized as the mobile robot finds goal locations and other unique locations, and are then linked as paths between goal locations, as goals are found during exploration. The place code creates a hybrid cognitive map of metric and topological data. In doing so, much less memory is needed to represent the robot's roaming environment, as compared to traditional mapping methods, such as occupancy grids. A comparison of the memory and processing savings are presented, as well as to the functional similarities of our design to the rodent's specialized navigation cells.