# Transfer Policies From Simulation to Reality[#](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/transfer-policies-from-simulation-to-reality/index.html.md#transfer-policies-from-simulation-to-reality 'Link to this heading')

Welcome to our module on _Transferring Robot Learning Policies from Simulation to Reality_, which we often refer to as **sim-to-real**.

While this module won’t focus specifically on Isaac Lab, the principles we’ll cover are important to understand before you start developing robot learning policies in Isaac Lab. Our goal in this module is to teach you about sim-to-real, including best practices and pitfalls to avoid.

## Learning Objectives[#](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/transfer-policies-from-simulation-to-reality/index.html.md#learning-objectives 'Link to this heading')

By the end of this module you’ll be able to:

- Understand the challenges of reinforcement learning in robotics.

- Explain the concept of the “reality gap” and its characteristics.

- Describe various techniques for bridging the reality gap.

- Evaluate strategies for leveraging privileged information in robot learning.

- Identify best practices for successful sim-to-real transfer in robotics applications.

Note

This module does not contain hands-on activities.

## Connect These Ideas to the SO-101 Task[#](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/transfer-policies-from-simulation-to-reality/index.html.md#connect-these-ideas-to-the-so-101-task 'Link to this heading')

The [Train a Robot Arm](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/train-a-robot-arm/index.html.md) module teaches the nominal vial-placement task. The tutorial also supplies a `sim2real` preset for the same state and camera tasks. It connects several ideas in this module:

- Both tasks randomize contact properties, vial mass, joint dynamics, encoder bias, and object placement. Train the teacher and student under matching physics.

- The camera task additionally varies appearance, camera mount, intrinsics, exposure, and blur. Two RGB frames and joint feedback remain the student’s inputs.

- Transfer training starts from home rather than all reset phases. It also changes the shoulder and elbow action scales and uses measured follower dynamics.

The tutorial sequence is nominal teacher training, teacher continuation under `sim2real`, then camera distillation under the same preset. Evaluate nominal and randomized policies separately: the transfer preset retains physical randomization during evaluation, so its scores answer a different question.