Fig. 9. The summation of feature attractors. (A) Hypothetical attractor in which retinal error is converted to a yaw force of the ant. (B) At three different goal locations (F, F_near and F_far) the stored retinal position of a local landmark, or, equivalently, equilibrium point of an associated attractor, will vary. (C,D) Weighted averages of attractors. If two attractors from snapshots taken close together (e.g. F and F_near) are averaged across retinal position (C) then the resulting combined attractor has a single equilibrium point that moves as the relative weightings change (solid circles). If the two attractors come from snapshots taken far apart (F and F_far) then the resulting combined attractor (D) has two stable equilibrium points (at F and F_far) that do not move as the weighting is changed. The zero crossings between the two equilibrium points (dashed portion of the attractor curve) are unstable as turns are away rather than towards the zero point. (E,F) Simulated routes to two feeders using spatially limited attractors of different weights. Further explanation is given in the text.