Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
, 201 (6), 585-97

How Variation in Head Pitch Could Affect Image Matching Algorithms for Ant Navigation

Affiliations

How Variation in Head Pitch Could Affect Image Matching Algorithms for Ant Navigation

Paul Ardin et al. J Comp Physiol A Neuroethol Sens Neural Behav Physiol.

Abstract

Desert ants are a model system for animal navigation, using visual memory to follow long routes across both sparse and cluttered environments. Most accounts of this behaviour assume retinotopic image matching, e.g. recovering heading direction by finding a minimum in the image difference function as the viewpoint rotates. But most models neglect the potential image distortion that could result from unstable head motion. We report that for ants running across a short section of natural substrate, the head pitch varies substantially: by over 20 degrees with no load; and 60 degrees when carrying a large food item. There is no evidence of head stabilisation. Using a realistic simulation of the ant's visual world, we demonstrate that this range of head pitch significantly degrades image matching. The effect of pitch variation can be ameliorated by a memory bank of densely sampled along a route so that an image sufficiently similar in pitch and location is available for comparison. However, with large pitch disturbance, inappropriate memories sampled at distant locations are often recalled and navigation along a route can be adversely affected. Ignoring images obtained at extreme pitches, or averaging images over several pitches, does not significantly improve performance.

Figures

Fig. 1
Fig. 1
Head movement in navigating ants. Ants were recorded walking across the normal substrate at the field site in Sevilla, under 5 load conditions (from top to bottom row): no load, small food item, medium food item, large food item and carrying another ant as shown by the images inserts. a The distribution of head angles over the entire recording. b The instantaneous head and body angle of the ant as it moved across the camera field of view (blue line—body angle and red line—head angle). Angle conventions used are shown by the insert. The head of the ant moves across a substantial pitch range and varies with load. The attitude of the head is only partially decoupled from the motion of the body indicating that there is no continuous stabilisation of gaze
Fig. 2
Fig. 2
The effect of pitch on the visual compass. a Left column shows an example of the panoramic views generated at the same location in our 3D world but with pitch angle varying from −40° to 40°, introducing substantial distortion of the view. a Right column shows the rIDFs calculated at 81 test when comparing the reference image (no pitch) to the same image at differing pitch angles (−40° to 40°). Blue lines are the individual rIDF values for each location (aligned so the correct direction is at 0°), while the red line shows the mean across all the locations tested. At 0° pitch, the reference heading (0°) is readily identified as a minimum, but as pitch increases it becomes increasingly difficult to reliably extract the correct heading. b Range of heading errors computed across 81 test locations under varying pitch (median are shown in red, and inter-quartile range by box). Error increases significantly when pitch exceeds ±15°. c Signal strength, i.e. the median rIDF value divided by the minimum rIDF, as pitch varies. The depth of the minimum, and hence its detectability, drops dramatically with head pitch
Fig. 3
Fig. 3
Approaches to reducing the impact of head pitch on visual compass. Left column (a, d, g, j, m) angular difference between the ant path and the direction selected at 81 test locations along a real ant route using five different memory and visual processing techniques: Closest Memory (orange boxes); Local Memory (green boxes); Full Memory (blue boxes); Limited Memory; and Average Image. Medians are shown by the black bar and the IQR by the box. From top to bottom are the combinations of pitch in memory and test: zero, small and large pitch in memory and test; followed by small pitch memory and large in test and vice versa. Middle column (b, e, h, k, n) angular error between the pitch angle at which the test image was sampled and the best match found in memory using the Closest, Local and Full Memory methods, respectively. Right column (c, f, i, l, o) distance in cm between the location at which the test image was sampled and the best match found in memory using the Closest, Local and Full Memory methods, respectively. Searching in local memories only allows a spatially close image to be found at a correlated pitch angle, hence good performance. However, when searching over the entire route memory, pitch matching dominates, leading to spatial mismatches and poor performance. Learning only at small pitch values <10° and averaging memory do not improve performance
Fig. 4
Fig. 4
Navigating real ant routes with realistic head movement. a Actual route followed by an ant through its cluttered environment in Sevilla, Spain, acting as ground truth. b, c Example paths followed by simulated ant following visual compass methodology and realistic head pitch for memory and route recapitulation (b large to small; c small to large). The iterative computation of direction can compensate for the errors introduced by pitch (b) but is still susceptible to failure (continuous loops in this case) (c)

Similar articles

See all similar articles

Cited by 7 articles

See all "Cited by" articles

References

    1. Baddeley B, Graham P, Philippides A, Husbands P. Holistic visual encoding of ant-like routes: navigation without waypoints. Adapt Behav. 2011;19:3–15. doi: 10.1177/1059712310395410. - DOI
    1. Baddeley B, Graham P, Husbands P, Philippides A. A model of ant route navigation driven by scene familiarity. PLoS Comput Biol. 2012;8:e1002336. doi: 10.1371/journal.pcbi.1002336. - DOI - PMC - PubMed
    1. Cartwright BA, Collett TS. How honey bees use landmarks to guide their return to a food source. Nature. 1982;295:560–564. doi: 10.1038/295560a0. - DOI
    1. Cartwright BA, Collett TS. Landmark learning in bees. J Comp Physiol A. 1983;151:521–543. doi: 10.1007/BF00605469. - DOI
    1. Cerdá X, Retana J. Links between Worker Polymorphism and Thermal Biology in a Thermophilic Ant Species. OIKOS. 1997;78(3):467–474. doi: 10.2307/3545608. - DOI
Feedback