Modern MuJoCo Robotics (2025)

A collection of robotics algorithms (Kinematics, Dynamics, Control) based on Pranav Bhounsule's bootcamp, fully refactored and modernized for the new DeepMind mujoco Python bindings.

The Project

The original course uses the deprecated mujoco-py wrapper which no longer works on modern Linux systems. This repository contains the modernized translations of those algorithms using the native DeepMind API.

Tech Stack

• Language: Python 3.10+
• Physics Engine: DeepMind MuJoCo (2.x/3.x)
• IDE: VS Code (Linux)

Examples Included

• Projectile Motion: Basic physics and simulation.

• 

• 2-Link Manipulator: Double pendulum dynamics and chaotic motion.

3. 3D Inverse Kinematics (Reaching)

• File: 3d_reach.py
• Description: Upgraded the system to a 3-DOF (Degree of Freedom) robotic arm capable of moving in 3D space (X, Y, Z).
• Math Used: Implements Jacobian Inverse Kinematics with Damping (mu_jacSite and numpy.linalg.pinv) to solve for joint angles based on a 3D target position.

4. Mobile Robotics (Differential Drive)

• File: car.py
• Description: A simulation of a 3-wheeled differential drive robot (like a Roomba or Pioneer robot).
• Physics Features:
  • Implements a Skid-Steering kinematic model.
  • Uses a Free Joint (6DOF) to allow navigation anywhere in the environment.
  • Features a friction-less caster wheel for stability.
• Control Logic: A state-machine loop that executes a "Figure 8" driving pattern (Forward -> Left Turn -> Right Turn -> Spin).

Demo

5. Computational Physics (Drag Force)

• File: 7_projectile_drag.py
• Description: A physics experiment comparing ideal motion vs. real-world air resistance.
• The Physics:
  • Red Ball: Vacuum physics (Ideal Parabola).
  • Blue Ball: Applies a custom drag force $F = -c \cdot v$ injected directly into the simulation loop.
• Implementation: Uses data.qfrc_applied to apply custom forces to specific degrees of freedom at every simulation step.

6. Contact Physics (Restitution)

• File: 8_bouncing.py
• Description: Demonstrates how to control the Coefficient of Restitution (Bounciness) using MuJoCo's solref parameters.
• The Setup (Left to Right):
  • Red: High Damping (Inelastic / Mud).
  • Blue: Zero Damping (Elastic / Superball).
  • Green: Medium Damping (Basketball).
• Key Concept: Shows how we can tune the simulation to model different material properties, from energy-absorbing clay to high-energy rubber.

Demo