2R Robot Arm Lab Experiment – Gravity Compensation and Reinforcement Learning

Mon, 08 Jun 2026 00:00:00 +0000

Motivation

The “Intelligent Robotics” practical course in the M-IAS master’s programme currently lacks a hands-on experiment bridging classical model-based control and modern learning-based methods. A 2R robot arm with torque control offers exactly this: students first experience the physical effects of gravity on the real system, then learn how an RL agent solves the same task in a data-driven way.

Work Packages

Hardware

Selection and commissioning of BLDC motors and motor drivers (e.g. ODrive / VESC)
Mechanical design and fabrication of arm segments (CAD, 3D printing or aluminium profile)
Integration of encoders and wiring

Software & Control

Implementation of forward and inverse kinematics
Model-based gravity compensation controller (real-time torque compensation)
ROS integration: topics, services and ros2_control

Reinforcement Learning

Building a simulation environment (Gazebo or MuJoCo) for RL training
Training an agent (e.g. PPO / SAC) on a target-reaching task
Sim-to-real transfer and evaluation on the real arm

Documentation

Creating lab instructions for use in the M-IAS practical course

Requirements

Knowledge of control theory and robotics (kinematics, dynamics)
Programming skills in Python and/or C++
Ideally experience with ROS
Interest in machine learning and reinforcement learning
Enthusiasm for hands-on hardware work and prototyping

This topic can be completed as a project or master’s thesis subject to agreement.

Supervision

Role	Name	E-Mail
Supervisor	Prof. Dr. Christian Pfitzner	christian.pfitzner@th-nuernberg.de

Gravity Compensation | Mobile Robotics Research Group — Prof. Dr. Christian Pfitzner

2R Robot Arm Lab Experiment – Gravity Compensation and Reinforcement Learning

Motivation

Work Packages

Requirements

Supervision