Intrinsic coordination of grid-cell modules overcomes self-localization deviations in mobile robots

FERNANDEZ LEON, J.A.; SARRAMONE, LUCA

Otro; IEEE Transactions (enviado - en revisión); 2023

Context. Bridging the gap between life sciences and engineering help us understanding the brain’s cognitive maps. One area for improvement in designing artificial autonomous vehicles is obtaining an accurate position estimation when only using environmental signals and having no access to a global positioning system. Most mammals contrarily show efficient strategies to solve this issue through grid and place cells in the brain´s navigation system. Problem. Little is known about the coordination of grid cell´s modules during spatial navigation and its effect on the animal´s position estimation. Hypothesis. In the absence of an explicit coordination mechanism, velocity coupling and spatial reference points in the face of place cell´s fields enable module coordination for better prediction. Methodology. Using mobile robot´s navigation, a model based on path integration was implemented, representing a realistic toroidal topology of population activity in grid cells. Place fields were modelled statically but randomly distributed across the arena for grid cell anchoring. Results. Different grid cell modules were tightly coordinated even when the position estimate within grid modules deviated substantially from the robot´s actual position during detrimental place anchoring. Our model predicts that modules´ coordination is required for the joint encoding of a coherent estimated trajectory. Discussions. Grid cells modules remain coordinated for self-localization despite deviations on position estimates. The bioinspired plausibility of our approach is discussed in the context of grid cell activities anchored to place fields for path integration in autonomous vehicles.