xai615-realworld

This repo provides minimal hands-on code for the class XAI615 in RealWorld Robot for Pick-n-Place Demo.

MuJoCo related code is employed from following DeepMind's repos:

• [MuJoCo] https://github.com/deepmind/mujoco
• [Perception] https://github.com/NVlabs/UnseenObjectClustering
• [UniversalRobot] https://github.com/ros-industrial/ur_modern_driver

Prerequisites

This repo is tested on following environment:

• Ubuntu: 20.04
• Python: 3.8.10
• mujoco: 2.3.2

Install dependencies

Mujoco Engine

pip install mujoco

pip install mujoco-python-viewer

RG2 - Gripper

pip install pymodbus==2.4.0

Descriptions

Below is a list of files and their descriptions:

• Realworld

  1. Gripper: OnRobot RG2 Gripper usage
  2. Get Topic: Get Topic /joint_states(=current joint value)
  3. Solve IK: Solve Inverse Kinematics
  4. Perception: Segmentation using UnseenObjectClustering method
  5. Pick-n-Place: Pick-n-Place Demo

• Simulation

  1. Solve inverse kinematics in various method with [General, Augmented, Nullspace projection, Repulse, RRT-Star]
  2. Trajectory planning method: [Quintic, Minimum Jerk, Linear movement]
  3. Velocity profile method: [Trapezoidal, s-Spline method]

• Final Demo

  1. Stack object: Stack one object over another.
  2. Pick-n-Place only one object: Pick-n-Place only one object via Finite-State-Machine.
  3. Pick-n-Place multiple objects: Pick-n-Place multiple objects.
  4. Pick-n-Place multiple objects only using mujoco : Pick-n-Place multiple objects using only mujoco simulator