Toward Computationally Efficient Path Generation and Push Planning for Robotic Nonprehensile Manipulation

Robots have been involved in our daily life more and more. This involvement comes along with a necessity for endowing robots to interact with real-world objects that are novel to them. Among many others, pushing is one of the important tasks for robotic arms (or manipulators) as it has been widely used in the pipeline of the more complicated tasks (e.g., grasping, singulation, reorienting, and similar other tasks.), In this paper, we present simple yet efficient path generation and push planning algorithms for pushing an object from a given point to a target point in the given grid map with static obstacles. Our path generation module aims to generate waypoints taking into account obstacles and quantizing them into circular regions in order to reduce the computational cost. The path is formed using the waypoints by minimizing the designed cost function. We also define a cost function for planning pushing steps aiming to both minimize the number of pushing actions and deviation from the path generated. We present comparative numerical results using simulated data. Our preliminary findings show promising results leaving room for further improvements.