Deep learning platform comparisons & a predictive model for real-time offline application

This paper investigates predictive modelling, a mathematical technique for analyzing data patterns to forecast future events and outcomes. As a key element of predictive analytics, predictive modelling leverages machine learning and data mining methods to anticipate activity, behavior, and trends based on both historical and current datasets. The study first compares various computing platforms suitable for deep learning applications. It then focuses on predicting the torque required by an exoskeleton system designed to assist users in lifting objects with minimal effort. Using sensor data embedded within the exoskeleton, torque predictions are performed in an offline, real-time setting, utilizing available feature inputs. The results demonstrate the effectiveness of predictive modelling for accurately estimating torque requirements based on historical sensor data. Finally, the paper discusses potential offline applications of the system, highlighting the broader implications of successful torque prediction for enhancing assistive technologies.