Vision accounts for over 80% of information intake, and innovations in artificial intelligence vision can bring fundamental transformations to embodied intelligence and other technologies. Dynamic flexible optical sensing (DFOS) technology achieves curvature-matched optical capture, focal switching without optical movement, and multi-plane light field acquisition by enabling on-demand curvature adjustment of the sensor itself. However, achieving repeatable and reversible deformation remains challenging for current electronic arrays. This paper proposes a DFOS system based on liquid integrated circuits (LICs), which integrates photosensitive ionic liquids (PILs) with stretchable liquid metal electrodes to enable active curvature modulation for geometrically adaptive optical sensing. The photosensitive ionic liquid exhibits selective color perception, reversible photo-thermal-electric responses, and excellent stability. This all-liquid architecture enables large-amplitude hydraulic-controlled curvature modulation (>31.9°), with a strain rate of approximately 1.62%/second, and demonstrates exceptional durability—remaining free of mechanical fatigue after 600 reversible cycles. With machine learning assistance, the system achieves multimodal signal acquisition for color, pattern, and distance recognition under concave, planar, and convex configurations through hydraulically controlled curvature modulation, maintaining high precision.

II. Background Introduction

Over 80% of the information humans perceive originates from the visual system, which also serves as the primary sensory system in embodied intelligence and related technologies. Historically, the development of artificial visual systems has primarily focused on lens-based imaging systems. For over seven decades, mainstream research has been dedicated to designing lens systems specifically adapted to static planar light sensors (Figure 1A). However, planar sensors inherently suffer from optical aberrations such as field curvature (Petzval surface) and vignetting, necessitating correction through complex, bulky, and costly multi-lens assemblies. Despite significant advancements in lens technology, optimization research aimed at overcoming these physical limitations at the sensor end remains relatively lagging.

Existing research has achieved biomimetic curved surface sensing (e.g., human eye or compound eye structures) by leveraging rigid semiconductors such as monocrystalline silicon and germanium, demonstrating sensing capabilities with wide field of view, deep depth of focus, and low aberration. Pioneering studies on zoom imagers have attempted to realize dynamic flexible optical sensing (DFOS), a technology that enables optical sensors to adjust curvature on demand. DFOS holds exceptional application value, not only reducing distortion and vignetting by minimizing field curvature aberration but also enabling lens-free focus switching through equivalent optical path length variations induced by curvature modulation. Additionally, DFOS can achieve simultaneous multi-focus plane image capture via deformation-driven optical adaptation.

Although these attempts achieved certain stretchability through geometric engineering, they remain constrained by the inherent mechanical limitations of solid-state optoelectronic materials and their intrinsic sensing principles. Current challenges include a narrow deformation range, insufficient mechanical durability, and a lack of color recognition capability. Traditional stretchable optoelectronic devices are prone to mechanical fatigue and structural resonance under dynamic conditions. Additionally, since optoelectronic materials themselves lack color discrimination ability, conventional planar optical sensors rely on Bayer filters. However, due to the rigid structure of these filters, integrating them into deformable optical systems presents fundamental difficulties.

Source: Sensor Expert Network