Computational Model of Motion Sickness Describing the Effects of Learning Exogenous Motion Dynamics.

The existing computational models used to estimate motion sickness are incapable of describing the fact that the predictability of motion patterns affects motion sickness. Therefore, the present study proposes a computational model to describe the effect of the predictability of dynamics or the pattern of motion stimuli on motion sickness. In the proposed model, a submodel - in which a recursive Gaussian process regression is used to represent human features of online learning and future prediction of motion dynamics - is combined with a conventional model of motion sickness based on an observer theory. A simulation experiment was conducted in which the proposed model predicted motion sickness caused by a 900 s horizontal movement. The movement was composed of a 9 m repetitive back-and-forth movement pattern with a pause. Regarding the motion condition, the direction and timing of the motion were varied as follows: (a) Predictable motion (M_P): the direction of the motion and duration of the pause were set to 8 s; (b) Motion with unpredicted direction (M_dU): the pause duration was fixed as in (M_P), but the motion direction was randomly determined; (c) Motion with unpredicted timing (M_tU): the motion direction was fixed as in (M_P), but the pause duration was randomly selected from 4 to 12 s. The results obtained using the proposed model demonstrated that the predicted motion sickness incidence for (M_P) was smaller than those for (M_dU) and (M_tU) and no considerable difference was found between M_dU and M_tU. This tendency agrees with the sickness patterns observed in a previous experimental study in which the human participants were subject to motion conditions similar to those used in our simulations. Moreover, no significant differences were found in the predicted motion sickness incidences at different conditions when the conventional model was used.