A Survey on Control Methods for Robotic Pushing with Single-point Contact

Abstract Pushing is an essential nonprehensile manipulation primitive, particularly useful when objects are too heavy, large, or irregularly shaped to be grasped. However, planar single-point contact pushing remains challenging due to underactuation, hybrid contact dynamics, and sensitivity to uncertainty in friction, mass distribution, and state estimation. This survey reviews control methods for planar single-point contact pushing from a control-oriented perspective, focusing on how control inputs are computed, what system knowledge is assumed, what sensing feedback is required, and which control objectives are addressed. The literature is organized into four main categories: reinforcement learning, data-driven, model predictive control and analytical controllers. By comparing these approaches, we highlight key trade-offs between model knowledge and generality, between computational cost and real-time responsiveness. Finally, we discuss current trends such as the increasing use of tactile feedback and online adaptation, and outline open challenges including standardized benchmarking and extending pushing control beyond quasi-static assumptions toward higher-velocity dynamic pushing.