Perancangan Robot Mainan Pintar Berbasis Sensor dan Teknologi Machine Learning

Abstract

This research aims to explore the potential of a smart toy robot in supporting the development of early childhood in the continuously evolving digital era. Evaluation is conducted through a series of tests including image classification testing, omnidirectional movement testing, as well as training and testing of sound classification models. Image classification testing is performed to assess the face detection ability at varying distances, where accuracy presentation is quantitatively measured. The test results indicate that the robot is capable of detecting faces with an average accuracy rate of 82.6% at a distance of 10 cm and 79.4% at a distance of 50 cm. Omnidirectional movement testing shows an average error rate of 1.2% with the highest variation in the direction of 90° at -5.33%. Meanwhile, training and testing of sound classification models yield a validation accuracy of 83.4% in identifying specific sounds. Technically, this robot effectively detects faces at different distances, performs omnidirectional movements with high precision, and identifies various sounds with good accuracy. The outcome of this research is the development of a smart toy robot that can serve as an educational tool and stimulation for the development of early childhood. The novelty of this research lies in the comprehensive integration of technology in solutions for children's development in the digital era. Thus, this study provides a significant contribution to understanding and utilizing technology for children's education more effectively and efficiently in this ever-changing era.

Keywords– robot; smart toy; machine learning; toddler motoric skills.