Coherently remapping toroidal cells but not Grid cells are responsible for path integration in virtual agents

Area of Science:

• Neuroscience
• Computational Neuroscience
• Cognitive Science

Background:

• Grid cells and place cells are believed to be crucial for spatial navigation and path integration.
• Concurrent remapping of these cells in new environments raises questions about their causal relationship.

Purpose of the Study:

• To investigate the dynamics of spatial cells, specifically grid and place cells, within a computational model across multiple environments.
• To explore the causal relationship between grid cell remapping and place cell activity during environmental changes.

Main Methods:

• Utilized a continuous attractor recurrent neural network to model spatial cell dynamics.
• Employed random resampling of place cell centers to simulate global remapping.
• Applied dimensionality reduction techniques to analyze cell activity patterns.

Main Results:

• A subset of cells exhibited a persistent toroidal structure across different environments, resembling band-like cells.
• Pruning studies indicated that these toroidal cells are critical for path integration, while grid cells were found not to be.
• The model's generalization boundaries were delineated when extended to numerous environments, revealing current limitations.

Conclusions:

• Toroidal cells play a more significant role in path integration than previously thought, challenging the established role of grid cells.
• Current computational models face challenges in accurately simulating spatial navigation across a large number of diverse environments.