[Skip to main content](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/transfer-policies-from-simulation-to-reality/05-bridging-the-gap-policy-robustness/01-leveraging-privileged-information.html.md#main-content) Back to top Ctrl+K [Getting Started With Isaac Lab](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/index.html.md) Choose version Search Ctrl+K - [YouTube](https://www.youtube.com/@NVIDIAOmniverse 'YouTube') - [Discord](https://discord.gg/nvidiaomniverse 'Discord'.md) - [Survey](https://surveys.hotjar.com/98510484-7ad7-4ddc-bfc4-1b7663827216.md?utm_source=getting-started-with-isaac-lab 'Survey') - [Home](https://docs.nvidia.com/learning/physical-ai/ 'Home') Search Ctrl+K [Getting Started With Isaac Lab](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/index.html.md) Choose version - [YouTube](https://www.youtube.com/@NVIDIAOmniverse 'YouTube') - [Discord](https://discord.gg/nvidiaomniverse 'Discord'.md) - [Survey](https://surveys.hotjar.com/98510484-7ad7-4ddc-bfc4-1b7663827216.md?utm_source=getting-started-with-isaac-lab 'Survey') - [Home](https://docs.nvidia.com/learning/physical-ai/ 'Home') Table of Contents Contents - [Robot Learning Fundamentals](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/robot-learning-fundamentals/index.html.md) - [Motivation and Definition](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/robot-learning-fundamentals/01-motivation-and-definition.html.md) - [Problem Specification](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/robot-learning-fundamentals/02-problem-specification.html.md) - [Forms of Learning](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/robot-learning-fundamentals/03-forms-of-learning.html.md) - [Reinforcement Learning](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/robot-learning-fundamentals/04-reinforcement-learning.html.md) - [Proximal Policy Optimization](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/robot-learning-fundamentals/05-proximal-policy-optimization.html.md) - [Reward Design and Evaluation](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/robot-learning-fundamentals/06-reward-design-and-policy-evaluation.html.md) - [Policy Distillation](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/robot-learning-fundamentals/07-policy-distillation.html.md) - [Simulation to Reality](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/robot-learning-fundamentals/08-simulation-to-reality.html.md) - [Robot Learning Workflows](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/robot-learning-fundamentals/09-robot-learning-workflows.html.md) - [Isaac Lab Basics](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/isaac-lab-basics/index.html.md) - [The Isaac Platform](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/isaac-lab-basics/01-the-isaac-platform.html.md) - [Project Generation](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/isaac-lab-basics/02-isaac-lab-projects.html.md) - [Development Workflow](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/isaac-lab-basics/03-development-workflow.html.md) - [The Isaac Lab CLI](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/isaac-lab-basics/04-isaac-lab-cli.html.md) - [Manager-Based Workflows](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/isaac-lab-basics/05-manager-based-workflows.html.md) - [The Cartpole Example](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/isaac-lab-basics/06-isaac-lab-cartpole.html.md) - [Cartpole Training](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/isaac-lab-basics/07-cartpole-training.html.md) - [Train a Robot Arm](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/train-a-robot-arm/index.html.md) - [Project Setup](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/train-a-robot-arm/01-project-setup.html.md) - [SO-101 Robot](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/train-a-robot-arm/02-so101-robot.html.md) - [Vial and Rack Scene](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/train-a-robot-arm/03-vial-and-rack.html.md) - [Running Simple Agents](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/train-a-robot-arm/04-simple-agents.html.md) - [Train a Teacher Policy](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/train-a-robot-arm/05-teacher-policy.html.md) - [Distill a Student Policy](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/train-a-robot-arm/06-distilled-policy.html.md) - [Managing Experiments](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/train-a-robot-arm/07-managing-experiments.html.md) - [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) - [Overview](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/transfer-policies-from-simulation-to-reality/index.html.md) - [The Successes of Reinforcement Learning](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/transfer-policies-from-simulation-to-reality/01-reinforcement-learning/index.html.md) - [Reinforcement Learning](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/transfer-policies-from-simulation-to-reality/01-reinforcement-learning/index.html.md) - [The Challenges of Reinforcement Learning](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/transfer-policies-from-simulation-to-reality/01-reinforcement-learning/01-the-challenges-of-reinforcement-learning.html.md) - [Simulation: A Game-Changer](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/transfer-policies-from-simulation-to-reality/01-reinforcement-learning/02-simulation.html.md) - [What Is the Reality Gap?](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/transfer-policies-from-simulation-to-reality/02-the-reality-gap/index.html.md) - [The Reality Gap](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/transfer-policies-from-simulation-to-reality/02-the-reality-gap/index.html.md) - [Visualizing the Reality Gap](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/transfer-policies-from-simulation-to-reality/02-the-reality-gap/01-visualizing-the-reality-gap.html.md) - [Bridging the Gap: Simulation Enhancement](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/transfer-policies-from-simulation-to-reality/03-bridging-the-gap-simulation-enhancement/index.html.md) - [Domain Randomization](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/transfer-policies-from-simulation-to-reality/03-bridging-the-gap-simulation-enhancement/index.html.md) - [Visual Domain Randomization](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/transfer-policies-from-simulation-to-reality/03-bridging-the-gap-simulation-enhancement/01-visual-domain-randomization.html.md) - [Depth Camera Randomization](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/transfer-policies-from-simulation-to-reality/03-bridging-the-gap-simulation-enhancement/02-depth-camera-randomization.html.md) - [Point Cloud Perturbations](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/transfer-policies-from-simulation-to-reality/03-bridging-the-gap-simulation-enhancement/03-point-cloud-perturbations.html.md) - [Bridging the Gap: Real-World Data Integration](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/transfer-policies-from-simulation-to-reality/04-bridging-the-gap-real-world-data-integration/index.html.md) - [System Identification](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/transfer-policies-from-simulation-to-reality/04-bridging-the-gap-real-world-data-integration/index.html.md) - [Actuator Modeling](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/transfer-policies-from-simulation-to-reality/04-bridging-the-gap-real-world-data-integration/01-actuator-modeling.html.md) - [Digital Twins](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/transfer-policies-from-simulation-to-reality/04-bridging-the-gap-real-world-data-integration/02-digital-twins.html.md) - [Neural Rendering](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/transfer-policies-from-simulation-to-reality/04-bridging-the-gap-real-world-data-integration/03-neural-rendering.html.md) - [World Models](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/transfer-policies-from-simulation-to-reality/04-bridging-the-gap-real-world-data-integration/04-world-models.html.md) - [Regularization](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/transfer-policies-from-simulation-to-reality/05-bridging-the-gap-policy-robustness/index.html.md) - [Regularization](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/transfer-policies-from-simulation-to-reality/05-bridging-the-gap-policy-robustness/index.html.md) - [Leveraging Privileged Information](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/transfer-policies-from-simulation-to-reality/05-bridging-the-gap-policy-robustness/01-leveraging-privileged-information.html.md#) - [Considerations When Bridging the Reality Gap](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/transfer-policies-from-simulation-to-reality/06-conclusion/index.html.md) - [Considerations](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/transfer-policies-from-simulation-to-reality/06-conclusion/index.html.md) - [Tips for Sim-to-Real](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/transfer-policies-from-simulation-to-reality/06-conclusion/01-tips-for-sim-to-real.html.md) - [Review](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/transfer-policies-from-simulation-to-reality/06-conclusion/02-review.html.md) - [Quiz](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/transfer-policies-from-simulation-to-reality/06-conclusion/03-quiz.html.md) Resources [Physical AI Learning Hub](https://docs.nvidia.com/learning/physical-ai/) [Developer Forums](https://forums.developer.nvidia.com/c/omniverse/300) [Community Discord](https://discord.gg/nvidiaomniverse.md) - [Home](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/index.html.md) - [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) - [Regularization](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/transfer-policies-from-simulation-to-reality/05-bridging-the-gap-policy-robustness/index.html.md) - Leveraging Privileged Information # Leveraging Privileged Information[#](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/transfer-policies-from-simulation-to-reality/05-bridging-the-gap-policy-robustness/01-leveraging-privileged-information.html.md#leveraging-privileged-information 'Link to this heading') The last topic we need to cover to finish our discussion on sim-to-real is privileged information, a concept we introduced at the beginning of this module. Leveraging privileged information is a crucial topic in robotics and simulation. Let’s explore what it means and how it’s used. **Privileged information** refers to data that’s easily accessible in simulation but not readily available on a real robot. This can include: - Ground friction coefficients - Scene semantics - Exact object poses This additional information we can get from the simulation can be leveraged to make your systems even better. There are a couple different ways we can leverage this information: asymmetric actor-critic and the teacher-student approach. ## Asymmetric Actor-Critic[#](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/transfer-policies-from-simulation-to-reality/05-bridging-the-gap-policy-robustness/01-leveraging-privileged-information.html.md#asymmetric-actor-critic 'Link to this heading') In actor-critic methods, we train two networks: a value function (critic) and a policy (actor). The value function is necessary to update the policy during training but isn’t used during deployment. This allows us to give the value function access to privileged information, speeding up convergence. For instance, when training a grasping task using images, the policy receives the image input. However, the value function can access ground truth information about the cube it’s trying to grasp. This approach simplifies training with images and accelerates the learning process. ## Policy Distillation / Teacher-Student Approach[#](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/transfer-policies-from-simulation-to-reality/05-bridging-the-gap-policy-robustness/01-leveraging-privileged-information.html.md#policy-distillation-teacher-student-approach 'Link to this heading') This widely-used method involves three stages: 1. Train a teacher with access to privileged information 2. Train a student without privileged information to imitate the teacher 3. Deploy the student on the robot Let’s look at an example from the DextrAH-G project: **Stage 1**: The teacher is trained using reinforcement learning. It has access to privileged information like the full state of the robot and the ground truth pose of objects. **Stage 2**: The student is trained to imitate the teacher’s actions using only the information available to a real robot, such as depth images. **Stage 3**: The student policy is deployed on the real robot, as it doesn’t rely on privileged information. This approach allows us to leverage the benefits of privileged information during training while still producing a policy that can function in the real world with limited information. By using these techniques, we can create more efficient and effective robotic systems that bridge the gap between simulation and reality. [previous\\ \\ Regularization](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/transfer-policies-from-simulation-to-reality/05-bridging-the-gap-policy-robustness/index.html 'previous page') [next\\ \\ Considerations When Bridging the Reality Gap](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/transfer-policies-from-simulation-to-reality/06-conclusion/index.html 'next page') On this page - [Asymmetric Actor-Critic](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/transfer-policies-from-simulation-to-reality/05-bridging-the-gap-policy-robustness/01-leveraging-privileged-information.html.md#asymmetric-actor-critic) - [Policy Distillation / Teacher-Student Approach](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/transfer-policies-from-simulation-to-reality/05-bridging-the-gap-policy-robustness/01-leveraging-privileged-information.html.md#policy-distillation-teacher-student-approach) [View this page as Markdown](https://docs.nvidia.com/learning/physical-ai/getting-started-with-isaac-lab/latest/transfer-policies-from-simulation-to-reality/05-bridging-the-gap-policy-robustness/01-leveraging-privileged-information.html.md 'View this page as clean Markdown for LLMs and AI agents') [NVIDIA](https://www.nvidia.com) [Privacy Policy](https://www.nvidia.com/en-us/about-nvidia/privacy-policy.md/) | [Manage My Privacy](https://www.nvidia.com/en-us/about-nvidia/privacy-center.md/) | [Do Not Sell or Share My Data](https://www.nvidia.com/en-us/preferences/start.md/) | [Terms of Service](https://www.nvidia.com/en-us/about-nvidia/terms-of-service.md/) | [Accessibility](https://www.nvidia.com/en-us/about-nvidia/accessibility.md/) | [Corporate Policies](https://www.nvidia.com/en-us/about-nvidia/company-policies.md/) | [Product Security](https://www.nvidia.com/en-us/product-security.md/) | [Contact](https://www.nvidia.com/en-us/contact/) Copyright © 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved.