Abstract
When humans walk along a path with predefined foot targets, visual information about the location of an upcoming target is maximally important during the latter half of the preceding step (Matthis et al., 2017). However, strategies for visual sampling and control may differ when footholds are not predefined. When potential footholds are all regions that do not contain obstacles (e.g., a trail with small puddles), choices have to be made about where to place the feet. Furthermore, when obstacles are elevated, walkers may try to find footholds for one foot that allow the other foot to swing between rather than over obstacles (Fajen et al., VSS 2018). This may require visual information from a wider span of the upcoming terrain. In this study, subjects walked along a short path while stepping over or around virtual obstacles, which were projected onto the floor by a 3d-capable projector (Binaee & Diaz, 2018) and appeared as an array of randomly distributed blocks. There was also a condition in which obstacles were flat. The terrain was visible only when it fell within a donut-shaped region centered on the subject. We manipulated both the inner radius and the span (difference between inner and outer radii) of the visibility donut. With flat obstacles, collisions were infrequent and minimally affected by the visibility manipulations. When obstacles were 3d, performance was comparable to full vision only when the visibility donut spanned one to three step-lengths ahead. When the visibility span was smaller than two step-lengths or when obstacles disappeared more than one step-length in advance, collisions increased. The findings suggest that when obstacles are elevated, the region of visible terrain must be large enough to span two potential footholds. This may allow walkers to evaluate potential footholds in pairs, which could facilitate safer and more efficient locomotion.
Acknowledgement: NSF 1431078