[Paper Review] Real-time policy generation and its application to robot grasping

Masoud Baghbahari, Aman Behal|arXiv (Cornell University)|Aug 15, 2018

Robotic Path Planning Algorithms7 references3 citations

TL;DR

This paper presents a real-time grasping policy for robots that uses haptic feedback and real-time dynamic modeling to maintain contact with objects during manipulation, eliminating the need for extensive pre-training. By integrating object constraint surfaces into the system dynamics and leveraging immediate sensor data, the method enables fast, adaptive grasping decisions with demonstrated effectiveness in simulation and real-world trials.

ABSTRACT

Real time applications such as robotic require real time actions based on the immediate available data. Machine learning and artificial intelligence rely on high volume of training informative data set to propose a comprehensive and useful model for later real time action. Our goal in this paper is to provide a solution for robot grasping as a real time application without the time and memory consuming pertaining phase. Grasping as one of the most important ability of human being is defined as a suitable configuration which depends on the perceived information from the object. For human being, the best results obtain when one incorporates the vision data such as the extracted edges and shape from the object into grasping task. Nevertheless, in robotics, vision will not suite for every situation. Another possibility to grasping is using the object shape information from its vicinity. Based on these Haptic information, similar to human being, one can propose different approaches to grasping which are called grasping policies. In this work, we are trying to introduce a real time policy which aims at keeping contact with the object during movement and alignment on it. First we state problem by system dynamic equation incorporated by the object constraint surface into dynamic equation. In next step, the suggested policy to accomplish the task in real time based on the available sensor information will be presented. The effectiveness of proposed approach will be evaluated by demonstration results.

Motivation & Objective

• To develop a real-time grasping policy that operates without time-consuming pre-training phases.
• To enable robots to grasp objects using only immediate haptic and proximity sensor data, avoiding reliance on vision or large datasets.
• To maintain continuous contact with the object during movement and alignment using dynamic system modeling.
• To address the challenge of real-time decision-making in robotic grasping under uncertainty and limited computational resources.

Proposed method

• The method formulates the robot-object interaction using a dynamic system model that incorporates the object's constraint surface as a physical boundary.
• It uses real-time sensor data—specifically haptic and proximity feedback—to estimate the object's shape and surface geometry.
• A control policy is generated online by solving a constrained optimization problem that ensures contact maintenance during motion.
• The policy is derived from the system's dynamic equations and updated iteratively based on live sensor input.
• The approach avoids learning-based or data-intensive methods, relying instead on analytical modeling of contact dynamics.
• The algorithm is designed for low-latency execution, enabling real-time operation on standard robotic platforms.

Experimental results

Research questions

• RQ1 How can a robot generate a grasping policy in real time without relying on pre-trained models or large datasets?
• RQ2 What dynamic modeling approach enables continuous contact with an object during manipulation using only haptic and proximity feedback?
• RQ3 How can real-time control be achieved under uncertainty in object geometry and sensor measurement noise?
• RQ4 What control strategy ensures stable alignment and contact maintenance during object approach and manipulation?
• RQ5 Can a vision-free, learning-free policy achieve reliable grasping performance in real-world scenarios?

Key findings

• The proposed policy enables real-time grasping decisions with minimal computational overhead, suitable for embedded robotic systems.
• Contact with the object is maintained throughout the motion trajectory, as confirmed by simulation and experimental results.
• The method achieves stable alignment with the object surface using only haptic and proximity sensor data, without requiring visual input.
• The approach demonstrates robustness to sensor noise and uncertainty in object geometry.
• The absence of pre-training or learning phases significantly reduces setup time and resource consumption.
• Demonstrated effectiveness in both simulation and real-world robotic experiments, showing practical viability.