Personalized Trajectory Planning Algorithm Considering Passenger Motion Sickness Level

Autonomous vehicle, though rapidly developing, still faces several challenges, one of which is motion sickness. It is necessary to mitigate motion sickness without sacrificing driving safety. Even if experiencing the same vibrational stimulus, people experience motion sickness differently. However, in the field of motion planning that considers motion sickness, current studies are mostly focusing on how to reduce the discomfort index in general ways, i.e. not considering for any specific passenger. Therefore, to exploit further potential of sickness mitigation, two important issues need to be solved, 1) how to predict the motion sickness level; 2) how to plan a sickness-less trajectory. To this end, firstly we construct a radial basis function neural network model to predict the motion sickness level of passenger, based on the extensive experimental data that we accumulated previously. Then, we integrate this passenger-specific model into trajectory planning in autonomous vehicles, which is formulated as an optimal control problem. Finally, two typical scenarios prone to evoking motion sickness have been simulated to validate the proposed algorithm, including the lane change while decelerating and the longitudinal accelerating scenarios. Results demonstrate that our algorithm can generate safe and comfortable trajectories that are customizable according to passenger's specific susceptibility to motion sickness. Particularly, for those highly susceptible to motion sickness, our algorithm can reduce their motion sickness degree by 21.55% and 15.68% in the two above scenarios, respectively, compared with the conventional planning algorithm based on polynomial.