Modular Optimization-based Inverse Kinematics of a UR5e Arm in Mujoco

A modular optimization framework to control the end effector of an industrial robotic arm solely based on the euclidean norm between goal pose and the actual pose of the end effector -- IK agnostic control (no need for IK equations)
Inspired from »

Control Methodology

The end effector of the UR5e robotic arm is controlled using local optimization techniques. The control approach utilizes:

• The scipy.minimize library for optimization.
• A custom gradient descent algorithm based on the Jacobian transpose method.

Tools involved

• Mujoco (latest)
• Python - scipy.optimize and numpy libraries

Getting Started

Steps to setup the framework in your local machine.

Prerequisites

Install the MuJoCo and SciPy libraries using pip to ensure the code runs without errors.

• pip

  python3 -m pip install mujoco scipy numpy

Installation

Considering that you have the prerequisites

1. Clone the repo

   git clone https://github.com/Poorna-Sasank/mujoco_ik_simulation.git

2. Navigate to optim_kinematic_control folder

   cd path/to/mujoco_ik_simulation/optim_kinematic_control

3. Choose your desired optimization framework and run the main code :)
   
   • You can choose your optimization framework in the main code itself

(back to top)

Usage

Should you require any further clarifications on mujoco or optimization techniques please refer to their respective documentations

For mujoco, please refer to the Mujoco Documentation
To learn about optimization methods, there are numerous resources available; I recommend starting with 3B1B

(back to top)