Smart Grasping for 2-Finger Grippers: A Deep Learning Approach for Autonomous Object Manipulation with Collaborative Robots

Project summary

This project aims to develop a robust, vision-driven software extension for 2-finger grippers which uses RGB-D data to enable precise, multi-modal grasping of objects with varying geometries. By integrating semantic visual data with geometric depth information, the system learns from human demonstrations to execute smooth, reactive trajectories on the Franka FR3 cobot, offering a scalable and intuitive solution for intelligent object manipulation in both simulated and real-world environments.