Abstract
Path integration is the updating of one's position and orientation during walking. Previously, we tested models of path integration in a triangle completion task (VSS 2008); here we attempt to predict performance on multi-segment paths. To perform such a homing task, estimates of the distances traveled and the angles turned must be combined to determine a return path to the home position. Large systematic errors are observed in triangle completion, which could be due to (i) encoding error in estimating the distances and angles traveled, (ii) integration error in combining these estimates to determine the homebound path, or (iii) execution error in turning and walking the homebound trajectory. Previously, we predicted triangle completion from performance on distance and angle reproduction tasks. Monte Carlo simulations of individual data revealed that a model based solely on encoding error (Fujita et al., 1993) does not account for errors in triangle completion. In contrast, models that include execution error predict triangle completion much more closely. In the present study, we test model predictions on more complicated paths. Participants wear a head-mounted display (60° H × 47° V) while walking in virtual hallways; head position is recorded using an inertial/ultrasonic tracker (70 ms latency). They walk a prescribed outbound path, then they turn and walk directly back to the remembered starting position in a new hallway determined by their heading. We varied the number of segments in the outbound path (2, 3, 4) while holding outbound path length constant (10 m), and varied outbound path length (10, 14, 18 m) while holding the number of outbound segments constant (3). As segment number or path length increases, so does the cumulative homing error. Model simulations demonstrate that execution error is a major factor determining path integration performance.
Supported by NSF BCS-0214383 and NASA/RI Space Grant.