Multifunctional Transition Metal Dichalcogenide‐Based Electronic Skin for AI‐Enhanced Multimodal Human–Machine Interaction and Robotic Control

Wearable electronic skin (e‐skin) provides a transformative approach to human–machine interaction by seamlessly integrating physiological sensing and environmental perception with robotic control. Here, we present a multifunctional e‐skin platform in which two‐dimensional transition‐metal dichalcogenide MoSe 2 supports NO 2 sensing, photodetection, and FET operation, whereas Au‐based structures enable surface electromyography (sEMG) acquisition and temperature sensing. Representative device performance included NO 2 sensing over 0.5–10 ppm, a photodetector on/off ratio exceeding 100, and a temperature sensitivity of 2.2 Ω °C − 1 . By integrating sEMG‐based muscle activity recognition with voice commands and a large language model‐assisted task planning module, we establish a multimodal human–machine interaction framework that combines high‐level semantic instructions with low‐level sEMG‐derived motor skills for intuitive robotic task execution. The proposed multimodal human–machine interaction framework enables robotic tasks, including object retrieval, object manipulation, and quadruped locomotion, by combining voice commands with sEMG‐based control, thereby facilitating the intuitive translation of human intentions into robotic actions. Using a six‐gesture sEMG dataset collected from one participant over 10 acquisition rounds, the e‐skin system was evaluated using 10‐fold round‐wise cross‐validation, with eight rounds used for training, one randomly assigned round for validation, and one round for testing in each fold. The GRU classifier achieved a mean fold accuracy of 97.88% ± 1.50% (population standard deviation), a pooled balanced accuracy of 97.91%, a Cohen's κ of 0.9747, and a macro‐F1 score of 97.92%, and was successfully integrated into real‐time robotic control applications. This work establishes a scalable foundation for next‐generation assistive robotics, teleoperation, and industrial automation, where seamless human‐robot collaboration is critical.