Adimen Artifiziala Erosotasunera Bideratuta: Zorabioaren Iragarpena Sistema Txertatu batean

Abstract

This work explores the development of a motion sickness prediction system on an embedded platform. The system is based on the ESP32 microcontroller and acquires real-time data from the vehicle’s CAN bus, GPS, and IMU sensors. The acquired data is processed using a CatBoost machine learning model to predict the probability of motion sickness. The model is optimized for efficient implementation on resource-constrained embedded systems, employing dynamic model management and the BIN format. The results demonstrate that the system is effective, achieving fast inference times and optimized memory usage. Future work will focus on the integration of actuators and the improvement of model performance.