Spatial disorientation in gondola centrifuges predicted by the form of motion as a whole in 3-D

Abstract

Introduction: During a coordinated turn, subjects can misperceive tilts. Subjects accelerating in tilting-gondola centrifuges without external visual reference underestimate the roll angle, and underestimate more when backward-facing than when forward-facing. In addition, during centrifuge deceleration, the perception of pitch can include tumble while paradoxically maintaining a fixed perceived pitch angle. The goal of the present research was to test two competing hypotheses: 1) that components of motion are perceived relatively independently and then combined to form a three-dimensional (3-D) perception; and 2) that perception is governed by familiarity of motions as a whole in three dimensions, with components depending more strongly on the overall shape of the motion.

Methods: Published experimental data from existing tilting-gondola centrifuge studies were used. The two hypotheses were implemented formally in computer models, and centrifuge acceleration and deceleration were simulated.

Results: The second, whole-motion oriented hypothesis better predicted subjects' perceptions, including the forward-backward asymmetry and the paradoxical tumble upon deceleration. The predominant stimulus at the beginning of the motion and the familiarity of centripetal acceleration were important factors.

Conclusion: Three-dimensional perception is better predicted by taking into account familiarity with the form of 3-D motion.

Figure 1

The two models. All vectors are in three dimensions and specified in coordinates aligned with the head (see text). (A) Componentwise Model. Inputs are angular acceleration and GIA. From angular acceleration, angular velocity is computed with decay time constant τ_a. From the angular velocity and the GIA, orientation is computed by the laws of physics (using a cross product) from the angular velocity but modified by a tendency with time constant τ_t toward vertical being aligned with the GIA. Perceived orientation is described by earth-upward vector g and by earth-horizontal vectors i and j representing fixed orthogonal compass directions, all specified in head coordinates in order to give their directions relative to the head. From the GIA and g, linear velocity is computed with a decay time constant of τ_l, and position is computed from linear velocity. Variables are as listed in the figure. (B) Whole-Motion Model. Inputs are angular acceleration and GIA. From the perceived earth-vertical component of angular acceleration, angular velocity (earth-vertical) is computed with decay time constant τ_a. From the angular velocity, heading (angle θ_h relative to a fixed compass direction) is computed. From the forward/backward perceived earth-horizontal component of the GIA (not necessarily equaling the noseward A_x if the head is pitched), linear velocity (forward or backward) is computed with decay time constant τ_l. This time constant is different for forward and backward motion, as explained in the text. Position is computed from linear velocity. From the angular and linear velocities, an expected centripetal acceleration and thus expected roll tilt of the GIA is computed and compared with the actual roll tilt of the GIA in order to determine the roll tilt angle, θ_r. The pitch angle θ_p is computed from the portion of the forward/backward GIA not used for linear acceleration, i.e. from the time-constant portion of the linear velocity differential equation.

Figure 2

Componentwise Model results for centrifuge acceleration. (A) Time course of predicted perceived roll and pitch for the 10s of centrifuge acceleration, for both forward-facing and backward-facing orientations, which gave the same results. Negative roll is toward the centrifuge axis, and negative pitch is away from the direction of motion. The dotted line shows actual roll of the subject. Experimental means for roll are also shown at the endpoint of acceleration for forward-facing (FF) and backward-facing (BF) orientations. (B) Polyhedral “head” used for display of three-dimensional position and orientation in Figures 23 and 4. (C) Three-dimensional predicted perceived motion—top view—during forward-facing acceleration, shown in freeze-frame format with a polyhedral head every 1s, starting at (0,0). On the last polyhedron (upper right) the right face is slightly visible because of 21° roll tilt to the left, matching the graph in part (A). The motion is described in the text.

Figure 3

Whole-Motion Model results for centrifuge acceleration. (A) Time course of predicted perceived roll and pitch for the 10s of centrifuge acceleration, for both forward-facing (FF) and backward-facing (BF) orientations. Negative roll is toward the centrifuge axis, and negative pitch is away from the direction of motion. The dotted line shows actual roll of the subject. Experimental means for roll are also shown at the endpoint of the acceleration. (B) Three-dimensional predicted perceived motion during forward-facing acceleration, displayed using the same conventions as in Fig. 2C. (C) Three-dimensional predicted perceived motion during backward-facing acceleration, displayed using the same conventions as in Fig. 2C. The motions are described in the text.

Figure 4

Deceleration in a counterclockwise-rotating centrifuge. (A) The stimulus: Angular velocity components computed from the three-dimensional angular acceleration. Positive roll is rightward, positive pitch is forward, and positive yaw is leftward. (B) Axis of the angular velocity resulting from the stimulus, for the first six seconds, back view relative to the head. The head is shown upright simply for clarity to indicate the angular stimulus relative to the head, and because the subject begins the deceleration with a perception of no roll tilt. Arrow lengths represent the order in which the axes occur, with the longest vector first to highlight the initial stimulus the most. Angles from horizontal are listed. (C) Predicted perceived roll and pitch, which is paradoxical, as described in the text. The starting point for this graph was 22° perceived pitch back (during centrifuge constant velocity). The graph starting with 0° pitch back (not shown) had pitch curves almost identical to these after the first two seconds. This graph was generated from the version of the model that included both pitch and roll input; the version of the model with only pitch input gave a very similar graph but with roll angle remaining zero. Positive roll is rightward, and positive pitch is forward. (D) Three-dimensional predicted perceived motion (including position and orientation) during deceleration, as viewed from the right side of the subject, displayed using the same conventions as in Fig. 2C. The starting point for this graph was 22° perceived pitch back. The three-dimensional display with starting orientation 0° pitch back was similar (not shown).