Neuroscience AI & Continuous Attractor Networks

VectorHaSH: Grid Cell Navigation Arena

Move a virtual agent across a 2D spatial environment to visualize hexagonal grid cell firing fields generated by bio-inspired Continuous Attractor Networks (CAN).

Spatial Arena & Path Integration

[Click or drag cursor inside the arena to move the agent]

Grid Module Scale: λ = 30 cmPath Error: 0.02 m
Grid Cell Firing Rates
Attractor Bump Location

(x: 1.42m, y: -0.88m)

Orientation Tuning

60° Hexagonal Symmetric

Scaffold Precision

High Localization Accuracy

Bio-Inspired Navigation Scaffold Active

How Bio-Inspired Grid Cell Navigation Works

Developed as part of bio-inspired robotics research in Prof. Schwarz's Lab at McGill University (co-authoring RatSLAM 2.0 / VectorHaSH), this lab models how entorhinal cortex grid cells encode spatial metrics.

Continuous Attractor Networks (CAN)

Models a 2D neural manifold where localized activity "bumps" shift in response to velocity input (path integration), maintaining stable spatial representations even without visual landmarks.

Hexagonal Firing Patterns

Demonstrates how individual grid cells fire at periodic triangular nodes forming a 60° spatial lattice, providing an internal coordinate system for robotic SLAM and biological navigation.

Explore the complete PyTorch code and RatSLAM 2.0 VectorHaSH repository on GitHub:View Repository