Figure 1
From: Learning self-driven collective dynamics with graph networks
Simulation of collective motion at different densities and noises. The actual speed of the individual is indicated by a small arrow, and the tail of the arrow is the actual position of the particle. In each case, the number of particles is N = 300. (a) In the initial state of L = 7 and \(\eta \) = 2.0, the position and direction of the particles are random. (b) In the case of low density and low noise (L = 25, \(\eta \) = 0.1), the particles in the collective will aggregate after a period of time to form small clusters, where different colors indicate that the particles belong to different clusters. (c) For high-density and high-noise (L = 7, \(\eta \) = 2.0), the movement direction of the particles in the collective is random in a small range, but there is a certain correlation overall. (d) In the case of high density and low noise (L = 5, \(\eta \) = 0.1), the individuals in the collective move in the same direction. All results are obtained from simulations with a rate set to 0.03.
