A Rat Brain-Inspired Model for Quadruped Robot Localization Using Path Integration and Visual Landmarks

Exceptional rodent navigation relies on the integration of sensory and proprioceptive information within the entorhinal-hippocampal system. Inspired by neurophysiological studies, a brain-like localization model based on the rat brain’s cognitive mechanism is proposed. To enhance biomimicry, the quadruped robot is used to simulate the physical structure of a rat. Firstly, a path integral model based on a quadrupedal structure is constructed to solve for the robot’s position in space. Then an entorhinal-hippocampal neurocomputing model is constructed to further process the position information, so as to realize the neural representation of path integral results. Finally, the cumulative error in the movement process is corrected using landmark information. To verify the model, several experiments have been carried out in the 3-D simulation environment. Experimental results show that the proposed method not only can well simulate the process of path integration using proprioception in rats, but also can show excellent localization performance. Quantitatively, the quadruped path integration model maintains a low average trajectory error of 0.2132 m. The integration of visual landmarks further reduces this error by 90.06% and boosts the navigation convergence times by 69.23%. These research advance the research on environmental cognition and navigation in bionic robots.