Sensory feedback in a bump attractor model of path integration

Area of Science:

• Neuroscience
• Computational Neuroscience
• Cognitive Science

Background:

• Mammalian spatial navigation relies on integrating multiple sensory inputs, including landmark (allothetic) and self-motion (idiothetic) cues.
• Neural networks like place cells, grid cells, and head direction cells encode spatial position.
• Path integration, while crucial, can be prone to errors from internal network imperfections.

Purpose of the Study:

• To develop a computational model merging path integration with external sensory cue feedback.
• To investigate how synaptic asymmetry and heterogeneity in neural networks affect spatial coding.
• To determine optimal parameters for sensory cue feedback to correct path integration errors.

Main Methods:

• Developed a computational model based on a continuous bump attractor network.
• Utilized asymptotic analysis to simplify the model into a scalar equation.
• Simulated the impact of synaptic asymmetry, heterogeneity, and dynamic noise on path integration accuracy.

Main Results:

• Synaptic imperfections disrupt the path integration position code, leading to accumulated errors.
• External sensory cues act as a control signal to correct these path integration errors.
• Optimal strength and decay rates for the control signal were identified for periodic and random cue appearances, and for dynamic noise.

Conclusions:

• External sensory feedback is essential for robust spatial positioning, especially when internal neural mechanisms are imperfect.
• The study provides insights into the optimal integration of sensory information for accurate navigation.
• The findings have implications for understanding and modeling neural mechanisms of spatial memory and navigation.