Real-time sensing and processing of tactile information are essential to enhance the capability of artificial electronic skins (e-skins), enabling unprecedented intelligent applications in tactile exploration and object manipulation. However, conventional tactile e-skin systems typically execute redundant data transfer and conversion for decision making due to their physical separation between sensors and processing units, leading to high transmission latency and power consumption. Here, we report an in-sensor tactile computing system based on a flexible capacitive pressure sensor array. This system utilizes multiple connected sensor networks to execute in-situ analog multiplication and accumulation operations, achieving both tactile sensing and computing functionalities. We experimentally implemented the in-sensor tactile computing system for low-level tactile sensory processing tasks including noise reduction and edge detection. The consumed power for single sensing-computing operation is over 22 times lower than that of a conventional mixed electronic system. These results demonstrate that our capacitive in-sensor computing system paves a promising way for power-constrained applications such as robotics and human-machine interfaces.

Despite the effort of analog circuit design automation, currently complex analog circuit design still requires extensive manual iterations, making it labor intensive and time-consuming. Recently, reinforcement learning (RL) algorithms have been demonstrated successfully for the analog circuit design optimization. However, a robust and highly efficient RL method to design analog circuits with complex design space has not been fully explored yet. In this work, inspired by multiagent planning theory as well as human expert design practice, we propose a multiagent based RL (MA-RL) framework to tackle this issue. Particularly, we (i) partition the complex analog circuits into several sub-blocks based on topology information and effectively reduce the complexity of design search space; (ii) leverage MA-RL for the circuit optimization, where each agent corresponds to a single sub-block, and the interactions between agents delicately mimic the best design tradeoffs between circuit sub-blocks by human experts; (iii) introduce and compare three different multiagent RL algorithms and corresponding frameworks to demonstrate the effectiveness of the MA-RL method. (iv) employing twin-delayed techniques and proximal policy to further boost training stability and accomplish higher performances. (v) The impacts of different reward function definitions as well as different state settings of MA-RL agents are investigated to further improve the robustness of this framework. (vi) Experiments on three different complex analog circuit topologies (GBA, DLL and SAR ADC) and knowledge transfers between two technology nodes are demonstrated. It’s shown that MA-RL framework can achieve the best FoM for complex analog circuits’ design. This work shines the light for future large scale analog circuit system design automation.

Mid‐infrared (Mid‐IR) photodetection and imaging are pivotal across diverse applications, including remote sensing, communication, and spectral analysis. Among these, single‐pixel imaging technology is distinguished by its exceptional sensitivity, high resolution attainable through the sampling system, and economic efficiency. The quality of single‐pixel imaging primarily depends on the performance of the photodetector and the sampling system. Photodetectors based on black phosphorus (BP) exhibit low dark current, high specific detectivity ( D ^* ), and room‐temperature operability. Artificial intelligence (AI)‐assisted sampling systems feature efficient and intelligent data reconstruction capabilities. In this work, we demonstrate an AI‐driven black phosphorus (BP)/molybdenum disulfide (MoS ₂ )/hexagonal boron nitride (hBN) heterojunction for Mid‐IR photodetection and imaging. By optimizing the thickness of the heterojunction, the quality of the interface, and the AI algorithm, we achieved high‐performance Mid‐IR photodetection and imaging. Specifically, the photodetector has a responsivity of 0.25 A/W at a wavelength of 3,390 nm, an extremely high D ^* of 3.7 × 10 ⁹ Jones, a response speed as low as 7 ms, and after AI optimization, the image contrast ratio has been improved from 0.227 to 0.890. At the same time, the sampling rate requirement can be reduced to 25 %. Our research indicates that the efficient combination of BP heterojunction photodetectors and AI technology is expected to accelerate the development of Mid‐IR photodetectors and imaging systems.

Most greenhouse gases come from biological activities and industry which will lead to global warming and show an impact on human life. With the need of green transformation of the global economic structure and seeking for higher quality of human life, the detection and management of greenhouse gases, as well as most hazardous gases in the environment, are increasingly demanding. Applications in different fields require sensors that can detect gas volume fractions with magnitudes from 10 ^-9 to 10 ^-4 . Greenhouse gas detection plays an important role both in the agriculture and industry field. In this review, we first summarize the mechanism of several common gas detectors used currently. Then, the advantages of nanostructured gas sensors are discussed. Finally, the applications of infrared gas sensors based on nanophotonic devices are described in detail. This review has been an outlook on the future development of infrared gas sensors based on nanophotonic devices.

Very long wavelength infrared (VLWIR, 14–30 μm) detection is crucial for observing low-temperature objects, such as those in infrared astronomy and Earth remote sensing, where focal plane arrays (FPAs) with ultra-low dark current and broadband spectral response are essential. Silicon-based blocked impurity band (BIB) photodetectors are promising candidates due to their intrinsically low dark current and sensitivity in the VLWIR range. However, simultaneously achieving a broad spectral response from 14 to 30 μm and high pixel-to-pixel uniformity remains a significant challenge. In this work, we report a broadband VLWIR FPA based on phosphorus-doped silicon (Si:P) BIB photodetectors. We designed a Si:P BIB architecture and a readout integrated circuit optimized for deep cryogenic operation. As a result, we developed a 16 × 16 prototype FPA with a 50% cutoff wavelength extended to 32.2 μm. Operating at 4 K, the array exhibits a peak detectivity of ∼1 × 1011 cm Hz1/2·W−1 and a dark current below 40 fA. This work provides a solution for future broadband VLWIR detection on both space-based and ground-based platforms.