Machine learning-enhanced low-hysteresis conductive auxetic strain sensors with curved re-entrant honeycomb structures based on MXene/graphene for human rehabilitation training

Flexible wearable electronic devices offer significant promise for the next generation of human–machine interfaces; however, they face challenges in improving sensing accuracy, flexibility, and mechanical property. Current conductive nanocomposites often suffer from high hysteresis and limited deformation recovery, hindering sensor performance. In this paper, we develop a novel hybrid nanocomposite composed of Ti₃C₂T_x MXene, graphene nanoplatelet (GNP), and silicone elastomer (MGS) through a mold-blade casting process by leveraging the auxetic effect. Designed with a unique curved re-entrant honeycomb architecture and the synergistic effect of the silicone and MXene/GNP nanobridges, this MGS composite exhibits impressive properties—low hysteresis and a negative Poisson's ratio under strain, as confirmed by finite element analysis. These innovations endow the MGS composites with excellent conformability, low hysteresis, a negative Poisson's ratio (−0.89), and remarkable mechanical resilience (98.6%). Moreover, the MGS composite demonstrates promising capabilities as a strain sensor, displaying a broad linear detection range, moderate sensitivity, and long-term durability. When integrated with a machine learning-based motion recognition system, the MGS sensor can accurately classify different knee bending angles with a 98.55% accuracy rate and provides real-time feedback, underscoring its significant potential for future applications in wearable human–machine interaction and rehabilitation training.