Embodied intelligence has witnessed remarkable progress in recent years,driven by advances in computer vision, natural language processing, and therise of large-scale multimodal models. Among its core challenges, robotmanipulation stands out as a fundamental yet intricate problem, requiring theseamless integration of perception, planning, and control to enable interactionwithin diverse and unstructured environments. This survey presents acomprehensive overview of robotic manipulation, encompassing foundationalbackground, task-organized benchmarks and datasets, and a unified taxonomy ofexisting methods. We extend the classical division between high-level planningand low-level control by broadening high-level planning to include language,code, motion, affordance, and 3D representations, while introducing a newtaxonomy of low-level learning-based control grounded in training paradigmssuch as input modeling, latent learning, and policy learning. Furthermore, weprovide the first dedicated taxonomy of key bottlenecks, focusing on datacollection, utilization, and generalization, and conclude with an extensivereview of real-world applications. Compared with prior surveys, our work offersboth a broader scope and deeper insight, serving as an accessible roadmap fornewcomers and a structured reference for experienced researchers. All relatedresources, including research papers, open-source datasets, and projects, arecurated for the community athttps://github.com/BaiShuanghao/Awesome-Robotics-Manipulation.