junchengli1/Sim-Grasp

Sim-Grasp offers a simulation framework to generate synthetic data and train models for robotic two finger grasping in cluttered environments.

🚀 Sim-Grasp

Sim-Grasp Learning 6-DOF Grasp Policies for Cluttered Environments Using a Synthetic Benchmark.
Juncheng Li, David J. Cappelleri.
Paper | Project Website

Table of Contents

• Cluttered Environment Generation
• Grasping Candidates Annotation

Cluttered Environment Generation

For dataset generation and simulation environment, refer to our previous work, Sim-Suction, which supports the generation of cluttered environments using Isaac Sim simulator.

Grasping Candidates Annotation

Use the two_finger_sampling_angle_aug.py script to sample grasp candidates and check collisions.

Use the grasp_simulation.py script to perform dynamic evaluation.

Sim-GraspNet

Use code under Sim_GraspNet for model construction (Codes to be cleaned).

Sim-Grasp-Policy

Use code under Sim_Grasp-policy for grasping policy (Codes to be cleaned).