GeoGraspEvo: multi-fingered grasp pose estimation

Grasping objects is a simple task for humans, but transferring this ability to robots is much more complex. Existing methods can be grouped into two categories: analytical and data-driven. In recent years, due to the advancement in computing capabilities, the research community has shifted its focus to the second group of algorithms. In this article, we propose a novel grasping method called GeoGraspEvo. It belongs to the group of analytical methods and allows to obtain the best combination of grasping points for multi-fingered grippers, starting from three fingers. The procedure is explained in full, starting from gripper parameters and a Point Cloud (PC), through the extraction of the gripping areas and candidate points, to the creation of ranking functions. Experiments are conducted on a widely recognized dataset such as Yale-CMU-Berkeley (YCB), in both simulated and real-world scenarios, employing a three-fingered gripper to demonstrate the efficacy of the presented method. The results demonstrate a successful grasping rate of 91.46% in simulation and 81.46% in the real-world with different shaped objects, with an execution time of 120 ms and 83 ms, respectively. In light of these findings, we have successfully matched and enhanced several state-of-the-art methods in terms of accuracy and computational efficiency across both categories with only Central Processing Unit (CPU) capabilities. Code is available at the project website.

Keywords: Manipulation Planning, Grasping and Manipulation, Grasping

Recommended citation: Ignacio de Loyola Páez-Ubieta, Santiago Timoteo Puente, Daniel Frau-Alfaro (2025). "GeoGraspEvo: multi-fingered grasp pose estimation. " Journal of Intelligent and Robotic Systems (JINT). Under review