Flexible sensors are becoming essential components of wearable healthcare, robotics, human–machine interfaces, and intelligent systems, where devices must detect complex signals while adapting to bending, stretching, and changing environmental conditions. Yet improving one sensor component often creates new limitations elsewhere: highly sensitive materials may have a narrow detection range, flexible structures may compromise stability, and sophisticated algorithms cannot fully compensate for noisy hardware or incompatible materials. Researchers from Harbin Institute of Technology, the University of Jinan, and collaborating institutions present a comprehensive review proposing a new materials–structure–hardware–algorithm co-design paradigm. This integrated approach provides a framework for overcoming fragmented development and advancing flexible sensing toward more reliable, intelligent, and application-ready systems.
Why This Flexible-Sensing Strategy Matters
Flexible sensors must transform physical or chemical stimuli into reliable electrical signals under demanding operating conditions. Conventional development often optimizes materials, device structures, electronic circuits, and data-processing algorithms separately, overlooking their interdependence. As a result, improvements in individual components do not necessarily translate into better system performance. For example, increasing conductivity may reduce mechanical flexibility, while integrating multiple sensing modes can compromise the performance of individual channels. This review emphasizes that next-generation sensors require coordinated optimization across the entire signal-perception chain rather than isolated improvements in individual parameters.
Innovative Design and Mechanism
The review organizes flexible-sensor development into four interconnected levels. Materials determine the fundamental sensing response, conductivity, mechanical compliance, and environmental compatibility. Structural engineering enhances signal conversion through architectures such as microstructured surfaces, porous networks, and multilayer configurations. Hardware provides efficient signal acquisition, conditioning, and transmission while minimizing noise, latency, and power consumption. Finally, algorithms interpret complex signals, separate overlapping responses, and enable real-time multimodal recognition. The central principle is that each level should be designed with the requirements of the others in mind. By coordinating material properties, device architecture, circuit implementation, and computational processing, researchers can resolve competing demands and achieve system-level improvements that isolated optimization may not deliver.
Outstanding Performance
Representative studies reviewed in the article demonstrate the potential of this integrated approach. Interlocked micro-dome structures have improved pressure-sensing detection limits to 0.2 Pa, while optimized hardware designs can reduce system response latency to the millisecond level. In an embodied-intelligence application, a hybrid neural network combining transfer learning with temporal-feature extraction achieved 98.33% recognition accuracy across 14 gestures, with a latency of just 0.1 seconds. Another multimodal sensing system integrated pressure, temperature, thermal conductivity, and electrical conductivity measurements with a compact convolutional neural network, achieving 98.03% recognition accuracy across 20 different objects. These examples illustrate how coordinated optimization can improve not only sensitivity or recognition accuracy, but also response speed, integration density, and practical usability.
Applications and Future Outlook
The proposed framework offers guidance for flexible sensing systems spanning wearable health monitoring, prosthetics, embodied intelligence, smart homes, intelligent machinery, and human–computer interaction. Future research should prioritize multimodal integration, emerging functional materials, low-power and compact hardware, and algorithms informed by the physical properties of sensing devices. The review also identifies important challenges, including the need for standardized evaluation methods capable of capturing nonlinear interactions and trade-offs between different design levels. Ultimately, the transition from component-level innovation to holistic co-design could enable flexible sensors that are more adaptable, robust, and intelligent, helping bridge the gap between laboratory demonstrations and practical deployment.