Search

Motion sickness is known to be an important and common factor affecting passengers' ride experience. With the popularization of intelligent vehicles, the development of multi-modal interaction methods will provide a chance to solve this problem. Therefore, the detection and mitigation of motion sickness will be an important topic for the future development of intelligent vehicles. However, detecting motion sickness using physiological and subjective data obtained in an on-road driving scenario has rarely been studied in related research. This paper proposed a solution framework for the physiological-signal-based motion sickness detection in on-road driving scenarios. Then, according to the framework, an on-road driving experiment is conducted to gather real-time physiological data using wearable devices. 12 participants took part in these experiments and about 120 min of motion sickness data were generated for analysis. The feature extraction was performed to analyze and extract the important physiological features related to motion sickness in real-world scenarios. Then the features were split into several combination groups and the representative machine learning models were trained to compare the results when different combinations were input. These detection models provide reasonable and effective detection when using the limited kind of physiological data. This method proposed in this paper will benefit the application of the motion sickness detection model in real vehicles. Meanwhile, since the visual input signals are not included, the solution is privacy-protected. In the future, with this solution, vehicles will also be able to detect motion sickness levels in real-time through low-cost daily wearable devices, such as smartwatches and glasses. © 2022 IEEE.