Purchase this article with an account.
Jonathan Samir Matthis, Karl S Muller, Mary M Hayhoe; Retinal optic flow and the control of locomotion. Journal of Vision 2019;19(10):179. doi: https://doi.org/10.1167/19.10.179.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
A long history of research on the visual control of locomotion has explored the role of optic flow in the regulation and guidance of human walking. It is generally agreed that the head-centered Focus of Expansion (FoE) lies in a stable location in the walker’s direction of travel, and that humans use this feature to control heading. We used optic flow estimation algorithms to measure head-centered optic flow recorded from the head-mounted camera of a mobile eye tracker of subjects walking in real-world environments. Contrary to the traditional view, we found natural head oscillations during locomotion cause the FoE to move constantly at high velocities within the walker’s field of view. Thus strategies that suggest the FoE is used for heading could not be implemented on the basis of naturally occurring optic flow. In contrast, we found that retinal optic flow contains information that may be used for the control of locomotion. Fixation nulls motion at the fovea, resulting in regular patterns of outward flow across the retina that encode information about the walker’s movement relative to the fixated location. Analyzing retinal flow fields using the curl and divergence operators from vector calculus reveals features that are directly applicable for the control of locomotion. For instance, the sign of retinal curl (which corresponds clockwise/counter clockwise rotation) indicates whether the walker will pass to the left or right of the point they are fixating. In addition, the walkers’ instantaneous, world-centered velocity vector may be derived directly from the divergence across the retinal flow field. These features provide a much richer and more stable source of information for the control of locomotion than the FoE. Furthermore, this information can be extracted directly from retinotopic visual motion, so a coordinate transform into head-centered coordinates is unnecessary.
This PDF is available to Subscribers Only