Human olfactory sensors have a large variety of receptor cells that generate signature responses to various gaseous molecules. Ideally, artificial olfactory sensors should have arrays of diverse sensors. However, it is challenging to monolithically integrate large-scale arrays of different high-performance gas sensors. Here we report biomimetic olfactory chips that integrate nanotube sensor arrays on nanoporous substrates with up to 10,000 individually addressable sensors per chip. The range of sensors is achieved using an engineered material composition gradient. Supported by artificial intelligence, the chips offer a high sensitivity to various gases with excellent distinguishability for mixed gases and 24 distinct odours. We also show that the olfactory chips can be combined with vision sensors on a robot dog to create a system that can identify an object in a blind box.

人类嗅觉传感器具有多种受体细胞，可以对各种气体分子产生特征响应。理想情况下，人工嗅觉传感器应该具有不同传感器的阵列。然而，单片集成不同高性能气体传感器的大规模阵列具有挑战性。在这里，我们报告了仿生嗅觉芯片，该芯片将纳米管传感器阵列集成在纳米多孔基板上，每个芯片上有多达 10,000 个可单独寻址的传感器。传感器的范围是通过工程材料成分梯度来实现的。在人工智能的支持下，该芯片对各种气体具有很高的灵敏度，对混合气体和 24 种不同的气味具有出色的辨别能力。我们还展示了嗅觉芯片可以与机器狗上的视觉传感器相结合，创建一个可以识别盲盒中物体的系统。