Automatic strain sensor design via active learning and data augmentation for soft machines

Emerging soft machines require high-performance strain sensors to achieve closed-loop feedback control. Machine learning is a versatile tool to uncover complex correlations between fabrication recipes and sensor performance at the device level. Here a three-stage machine learning framework was realized for a high-accuracy prediction model capable of automating the design of strain sensors. First, a support-vector machine classifier was trained by using 351 compositions of various nanomaterials. Second, through 12 active learning loops, 125 strain sensors were stagewise fabricated to enrich the multidimensional dataset. Third, to address the challenge of data scarcity, data augmentation was implemented to synthesize >10,000 virtual data points, followed by genetic algorithm-based selection to optimize the model’s prediction accuracy. Several data-driven design rules for piezoresistive nanocomposites were generalized and validated by in situ microscopic studies. As final demonstrations, model-suggested strain sensors can be integrated into/onto various soft machines to endow them with real-time strain-sensing capabilities.