Precision-Focused Reinforcement Learning Model for Robotic Object Pushing

Abstract-- Non-prehensile manipulation, such as pushing objects to a desired target position, is an important skill for robots to assist humans in everyday situations. However, the task is challenging due to the large variety of objects with different and sometimes unknown physical properties, such as shape, size, mass, and friction. This can lead to the object overshooting its target position, requiring fast corrective movements of the robot around the object, especially in cases where objects need to be precisely pushed. Humans intuitively interact with objects in everyday situations, object pushing based on a recurrent neural network (RNN) often without explicitly planning or thinking about how and model predictive control (MPC) cannot properly switch objects will behave. Non-prehensile object manipulation is an pushing sides, i.e. the model is not able to perform corrective important skill for robots that are designed to assist humans. Additionally, the authors also train a RL agent This work focuses on object pushing, a sub class of robotic as a model-free baseline.