Abstract
When walking a slalom course of five goals, occluding the next one or two upcoming goals significantly disrupts the control of steering (Zhao & Warren, VSS 2011). Specifically, the variable error and undershooting of goal positions increases from the Block 1 to the Block 2 condition. The results are consistent with the view that human locomotion is normally controlled on-line based on current visual information. The steering dynamics model (Fajen & Warren, 2003) can account for these walking trajectories on the basis of visual information about the next 1.5 goals. Here we ask whether the same control mechanism can account for steering to occluded goals, by incorporating positional uncertainty in the spatial memory of goal positions. On each trial, spatial uncertainty was simulated by randomly sampling from a Gaussian distribution of remembered positions centered on the actual goal position, for each goal in a slalom. The sampled goal positions were fed into the steering dynamics model to generate a simulated trajectory for that trial. The SD of the Gaussian distribution was treated as a free parameter. The results indicate that the variable error in human trajectories can be reproduced with a constant positional uncertainty for the Block 1 condition (SD = 0.073m for slalom goals 2-5), and increasing positional uncertainty for the Block 2 condition (SD=0.090m for slalom goal 2, 0.276m for goals 3-5). The pattern of undershooting in the Block 2 condition also implies a bias in the spatial memory for goals 3-5. The results indicate that an on-line steering control model can also account for steering to occluded goals simply by incorporating spatial uncertainty and bias in remembered goal positions.
Meeting abstract presented at VSS 2012