Toward Automated Tomato Harvesting System: Integration of Haptic Based Piezoresistive Nanocomposite and Machine Learning

Abstract

Carbon nanotubes (CNT)/polydimethylsiloxane (PDMS) have been investigated as potential materials for tomato-harvesting applications. The current-voltage (I-V) and current time (I-t) properties, as well as tomato hardness measurement and support-vector machine learning, were used to determine the performance of the sensor with respect to sensitivity, response time, accuracy, and detection limit of the nanocomposite. The data suggested an accurate (± 5.2%) measurement in a low-weight region of tomato. Narrowing of the I-V hysteresis curve towards a higher weight region was observed as a result of the increase in electron pathways. The fabricated sensor displayed a higher sensitivity (15 mV / mu text{m} ) than the commercial sensor (1 mV / mu text{m} ). In addition, machine learning of the resistance-displacement curve data yielded an average accuracy level of 0.67 when tested using acquired data.