This survey reviews data-driven grasp synthesis methods for robotic grasping, categorizing approaches based on whether they synthesize grasps for known, familiar, or unknown objects. For known objects, methods rely on object recognition and pose estimation. For familiar objects, similarity matching to previously encountered objects is used. For unknown objects, feature extraction is key. The survey discusses analytic and data-driven approaches, highlighting their differences in grasp sampling, quality estimation, and representation. Analytic methods provide guarantees but rely on assumptions like simplified contact models. Data-driven methods use empirical data, often from human demonstrations or simulations, to rank grasp candidates. They are more flexible but lack guarantees. The survey also compares data-driven methods to classical analytic formulations and identifies challenges in transferring grasp experience between objects. It emphasizes the role of perception, object representation, and the need for robustness in unstructured environments. Data-driven approaches are evaluated in simulation, but their real-world effectiveness is questioned. The survey concludes that data-driven methods are promising but require further development to better model real-world dynamics and improve generalization.This survey reviews data-driven grasp synthesis methods for robotic grasping, categorizing approaches based on whether they synthesize grasps for known, familiar, or unknown objects. For known objects, methods rely on object recognition and pose estimation. For familiar objects, similarity matching to previously encountered objects is used. For unknown objects, feature extraction is key. The survey discusses analytic and data-driven approaches, highlighting their differences in grasp sampling, quality estimation, and representation. Analytic methods provide guarantees but rely on assumptions like simplified contact models. Data-driven methods use empirical data, often from human demonstrations or simulations, to rank grasp candidates. They are more flexible but lack guarantees. The survey also compares data-driven methods to classical analytic formulations and identifies challenges in transferring grasp experience between objects. It emphasizes the role of perception, object representation, and the need for robustness in unstructured environments. Data-driven approaches are evaluated in simulation, but their real-world effectiveness is questioned. The survey concludes that data-driven methods are promising but require further development to better model real-world dynamics and improve generalization.