Fig 1. Simple, two-layer ECM (episodic and compositional memory) network for navigating the one-dimensional environment of our locust simulation.

The first layer of nodes represent percepts, which encode the net relative flow of nearby conspecifics s ∈ {−2, −1, 0, +1, +2}, whereas the second layer represent actions, of turning (a = −1) or continuing in the same direction (a = +1).