Machine-learning assisted decoupling of strain and temperature via an antagonistic sensing ionic elastomer

Solvent-free ionic elastomer with orthogonal electrical response • High-performance sensing: GF of 8.24 and TCR of 2.88%/°C • Machine learning decouples R/C signals into strain and temperature (R 2  ≥ 0.90) • Demonstrates simultaneous motion and heat detection for firefighters The simultaneous perception of multiple physical stimuli is essential for intelligent soft robotics and personalized healthcare. However, the concurrent detection of strain and temperature is fundamentally hindered by signal crosstalk, which compromises accuracy and reliability. Existing solutions, often relying on heterogeneous integration or intricate circuitry, inevitably sacrifice mechanical stability and scalability. Here, we introduce a sensing paradigm that leverages intrinsic material properties to achieve physical signal orthogonality. We design a solvent-free ion-conductive elastomer that exhibits antagonistic electrical responses to strain and temperature: its resistance increases but capacitance decreases upon stretching, whereas both resistance and capacitance increase with heating. This inherent signal decoupling stems from distinct ion migration and interfacial polarization mechanisms under mechanical and thermal stimuli. To translate this orthogonal relationship into precise quantitative readings, we employ a random forest regression model that directly maps the coupled resistance-capacitance signals to strain and temperature values, bypassing the need for complex analytical inversion. Our sensor achieves high-fidelity decoding, with determination coefficients of 0.90 for strain and 0.94 for temperature. It also maintains robust performance over 300 stretching cycles. We further demonstrate its practical utility in a firefighter safety warning system that simultaneously monitors body motion and detects extreme heat exposure. This work establishes a material-centric and algorithm-powered roadmap for multi perception, paving the way for high-fidelity sensing in complex environments. Download: Download high-res image (297KB) Download: Download full-size image