Boundary properties affecting idiothetic localization.

<p>(<b>A</b>) Culling of pose hypotheses over 3 true steps (red pointer), starting with 20 possible directions and two possible positions (40 poses). Each pose hypothesis independently tracked the displacements (blue) using only idiothetic displacement information. Pose estimates which crossed a boundary (red crosses) were removed. Acceptable hypotheses (cyan dots) were cloned to maintain the total particle population constant. For clarity, no sensory noise was included in this illustration but was present in all simulations. (<b>B</b>) Arenas of equal area (top) in ascending order (bottom to top, left to right) of rotational asymmetry functions (lower left) averaged over 360°. The adjusted <i>I_p^*</i> following 48 minutes increased with mean asymmetry (lower right). Regular polygonal arenas are labelled by the number of edges. (<b>C</b>) <i>I_p</i> (median, IQR) and <i>V(θ)</i> are shown using traversable space (blue) or boundary crossing (red) to update the pose estimate in a circular arena with an asymmetric barrier. Simulated grid cell spikes using a discrete model which did (right) or did not (left) use boundary crossing (46–48 min). There was a uniform angular distribution in traversable space (dotted line) but not traversed positions (solid lines).