<p>Diffusion models have demonstrated outstanding performance in image and video generation over recent years. Their strong expressive power and stable training dynamics have recently motivated researchers to integrate them into robot learning. In this survey, we trace and synthesize the latest progress in diffusion-based robot learning. We begin with an in-depth overview of robotics research history and the fundamentals of diffusion models. Next, we systematically examine diffusion-based methods in robotics, including diffusion policy approaches and diffusion-based image and video synthesizer. For diffusion policy-based methods, we present a comprehensive, top-down analysis of related approaches. We then explore the application of diffusion-based image and video generation techniques in robotic tasks, covering areas such as planning and data collection. Following this, we provide a review of key benchmarks and datasets, then summarize the major objectives addressed by diffusion-based methods and analyze their common strategies. Finally, we discuss the current challenges and outline promising future directions for diffusion-based robotic learning. Through this survey, we aim to provide a thorough resource inspiring further research at the intersection of diffusion modeling and robotics.</p>