Self-powered sensing systems augmented with machine learning (ML) represent a path toward the large-scale deployment of the internet of things (IoT). With autonomous energy-harvesting techniques, intelligent systems can continuously generate data and process them to make informed decisions. The development of self-powered intelligent sensing systems will revolutionize the design and fabrication of sensors and pave the way for intelligent robots, digital health, and sustainable energy. However, challenges remain regarding stable power harvesting, seamless integration of ML, privacy, and ethical implications. In this review, we first present three self-powering principles for sensors and systems, including triboelectric, piezoelectric, and pyroelectric mechanisms. Then, we discuss the recent progress in applied ML techniques on self-powered sensors followed by a new paradigm of self-powered sensing systems with learning capability and their applications in different sectors. Finally, we share our outlook of potential research needs and challenges presented in ML-enabled self-powered sensing systems and conclude with a road map for future directions.

借助机器学习 (ML) 增强的自供电传感系统代表了大规模部署物联网 (IoT) 的途径。借助自主能量收集技术，智能系统可以不断生成数据并对其进行处理以做出明智的决策。自供电智能传感系统的发展将彻底改变传感器的设计和制造，并为智能机器人、数字健康和可持续能源铺平道路。然而，在稳定的电力收集、ML 的无缝集成、隐私和道德影响方面仍然存在挑战。在这篇综述中，我们首先介绍了传感器和系统的三种自供电原理，包括摩擦电、压电和热电机制。然后，我们讨论了在自供电传感器上应用 ML 技术的最新进展，随后是具有学习能力的自供电传感系统的新范式及其在不同领域的应用。最后，我们分享了我们对支持 ML 的自供电传感系统中的潜在研究需求和挑战的展望，并以未来方向的路线图结束。