Influence of spatial orientation and spatial visualization abilities on space teleoperation performance

Astronauts perform space teleoperation tasks with visual feedback from outside cameras. Individuals differ greatly in the ability to integrate camera views, understand the workspace, and ensure clearances between the robot arm and obstacles. We believe that these individual differences correlate with two known subcomponents of spatial intelligence: perspective-taking (PT) and spatial visualization (SV). A preliminary study [1] supports this hypothesis. We believe astronauts use PT (the ability to imagine an object from a different viewpoint) to integrate camera information into an environmentally-referenced frame defined by the arm control axes. In some cases, it may be easier to visualize the manipulation of the payload with respect to the robot arm itself, than to the environment. In that case, SV(i.e., the ability to mentally manipulate an object from an egocentric perspective) may be exploited. We measured the performance of 25 naive subjects who used hand-controllers to rotate and translate, and 3 environmentally-fixed camera views. These devices controlled a 2-boom, 6 degree-of-freedom virtually-simulated arm to perform pickup and docking subtasks. To challenge the subjects' spatial ability we introduced a wide separation between camera views for some tests, and misalignments between the translation control and the display reference frames. We used the Perspective-Taking Ability test (PTA) and the Purdue Spatial Visualizations Test: Visualization of Views (PSVT:V) to measure PT, and the Cube Comparisons test (CC) to assess SV. We concluded that PTA predicted performance on pickup and docking subtasks, but PSVT:V did not. CC scores correlated with those measures of performance that did not necessarily require PT. High perspective-taking scorers performed the pickup task significantly more efficiently than low, but not faster. In docking, however, they were both significantly faster and more accurate, collided less often, and docked more accurately. In both tasks they moved along only one axis at a time. High CC scorers docked significantly more accurately and rotated about fewer axes at any one time. Whenever we found a significant effect of PSVT:V on a dependent variable, we also found one for PTA; but not the reverse. We had expected higher PT scorers to perform better than others under the challenge of wider camera angles and greater control-display frame misalignments, but we could not demonstrate this. On average females were slower and had lower docking accuracy, an effect related, perhaps, to their lower spatial ability scores. This study of performance during the first two hours of teleoperation training may help define issues for future research. Thesis Supervisor: Charles M. Oman Title: Senior Lecturer