--- title: Using AI Agents to Teach Robots Like Microduck to Walk... description: This workflow shows how AI agents and NVIDIA's robotics tools train a robot like Microduck to walk and navigate obstacles in simulation. AI agents Hermes,... canonical: https://daily.dev/posts/using-ai-agents-to-teach-robots-like-microduck-to-walk-in-simulation-kvzhhdhme twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Using AI Agents to Teach Robots Like Microduck to Walk in Simulation | daily.dev og:description: This workflow shows how AI agents and NVIDIA's robotics tools train a robot like Microduck to walk and navigate obstacles in simulation. AI agents Hermes,... og:url: https://daily.dev/posts/using-ai-agents-to-teach-robots-like-microduck-to-walk-in-simulation-kvzhhdhme og:image: https://api.daily.dev/og/posts/kVZhhdhme.png og:image:alt: Using AI Agents to Teach Robots Like Microduck to Walk in Simulation og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Using AI Agents to Teach Robots Like Microduck to Walk in Simulation **[NVIDIA Developer](https://daily.dev/sources/nvidiadev)** · 10 min read · 0 upvotes · 0 comments ## Summary This workflow shows how AI agents and NVIDIA's robotics tools train a robot like Microduck to walk and navigate obstacles in simulation. AI agents Hermes, Nemotron, and Astra turn a text prompt into a project plan, generate a concept image, and build 3D assets with FreeCAD and Blender. Isaac Lab trains a locomotion policy with Newton physics, and the ramp and obstacle adaptation run used 4,096 parallel environments on an RTX PRO 6000 Blackwell GPU. Trained policies are then validated locally in Isaac Sim on an RTX Spark before being prepared for deployment on real hardware, using Better Actuator Models and domain randomization that can help narrow the sim-to-real gap. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How many parallel environments did NVIDIA use to train the Microduck ramp-traversal policy in Isaac Lab? The adaptation run used 4,096 parallel environments on an NVIDIA RTX PRO 6000 Blackwell Workstation Edition GPU. A locomotion policy pretrained in the Isaac Lab Microduck example environment received additional training for traversing ramps and uneven surfaces, using head-camera depth and robot-state observations as inputs. _daily.dev keeps engineers scaling RL training runs current on GPU benchmarks like this one._ ### What does NVIDIA's Better Actuator Models (BAM) simulate for robot sim-to-real transfer in Isaac Lab? Better Actuator Models (BAM) in Isaac Lab model servo response, including command delays, mechanical backlash (play in the drivetrain), and voltage drops as motors draw power from a shared battery. Matching these settings to real robot measurements, combined with domain randomization of physical parameters and sensor noise, helps narrow the gap between simulated and real robot behavior. _Teams closing the sim-to-real gap can track actuator-modeling techniques like this on daily.dev._ ### What is the SimReady Central Library and how many assets does it contain? SimReady Central Library is a free, ungated collection of more than 2,000 simulation-ready OpenUSD assets for robotics work. Reusing these assets alongside newly generated objects reduces the asset preparation needed to assemble environments like an obstacle course in Isaac Sim. _Developers assembling simulation scenes can follow asset-library resources like this on daily.dev._ ## Similar posts on daily.dev - [From Simulation to Production: How to Build Robots With AI](https://daily.dev/posts/from-simulation-to-production-how-to-build-robots-with-ai-yezga7ll3) · NVIDIA · 0 upvotes · 0 comments - [How to Use AI Agents to Prepare 3D Scenes for Simulation](https://daily.dev/posts/how-to-use-ai-agents-to-prepare-3d-scenes-for-simulation-fsrfgl8eh) · NVIDIA Developer · 0 upvotes · 0 comments - [How WPP accelerates humanoid robot training 10x with G4 VMs](https://daily.dev/posts/how-wpp-accelerates-humanoid-robot-training-10x-with-g4-vms-r55syypbi) · Google Cloud · 0 upvotes · 0 comments --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#robotics](https://daily.dev/tags/robotics), [#reinforcement-learning](https://daily.dev/tags/reinforcement-learning) [View this post on daily.dev](https://daily.dev/posts/using-ai-agents-to-teach-robots-like-microduck-to-walk-in-simulation-kvzhhdhme) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Using AI Agents to Teach Robots Like Microduck to Walk in Simulation","url":"https://daily.dev/posts/using-ai-agents-to-teach-robots-like-microduck-to-walk-in-simulation-kvzhhdhme","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/using-ai-agents-to-teach-robots-like-microduck-to-walk-in-simulation-kvzhhdhme"},"datePublished":"2026-10-23T15:01:03.123Z","dateModified":"2026-10-23T15:02:12.435Z","description":"This workflow shows how AI agents and NVIDIA's robotics tools train a robot like Microduck to walk and navigate obstacles in simulation. AI agents Hermes,...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/1594077a17d8b12021532708890b6237?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/1594077a17d8b12021532708890b6237?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"NVIDIA Developer","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"NVIDIA Developer","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/86e45aab42ba48ce83103d01b1119910","url":"https://daily.dev/sources/nvidiadev"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/using-ai-agents-to-teach-robots-like-microduck-to-walk-in-simulation-kvzhhdhme","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-agents,robotics,reinforcement-learning","timeRequired":"PT10M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"NVIDIA Developer","item":"https://daily.dev/sources/nvidiadev"},{"@type":"ListItem","position":3,"name":"Using AI Agents to Teach Robots Like Microduck to Walk in Simulation"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/using-ai-agents-to-teach-robots-like-microduck-to-walk-in-simulation-kvzhhdhme#faq","mainEntity":[{"@type":"Question","name":"How many parallel environments did NVIDIA use to train the Microduck ramp-traversal policy in Isaac Lab?","acceptedAnswer":{"@type":"Answer","text":"The adaptation run used 4,096 parallel environments on an NVIDIA RTX PRO 6000 Blackwell Workstation Edition GPU. A locomotion policy pretrained in the Isaac Lab Microduck example environment received additional training for traversing ramps and uneven surfaces, using head-camera depth and robot-state observations as inputs. daily.dev keeps engineers scaling RL training runs current on GPU benchmarks like this one."}},{"@type":"Question","name":"What does NVIDIA's Better Actuator Models (BAM) simulate for robot sim-to-real transfer in Isaac Lab?","acceptedAnswer":{"@type":"Answer","text":"Better Actuator Models (BAM) in Isaac Lab model servo response, including command delays, mechanical backlash (play in the drivetrain), and voltage drops as motors draw power from a shared battery. Matching these settings to real robot measurements, combined with domain randomization of physical parameters and sensor noise, helps narrow the gap between simulated and real robot behavior. Teams closing the sim-to-real gap can track actuator-modeling techniques like this on daily.dev."}},{"@type":"Question","name":"What is the SimReady Central Library and how many assets does it contain?","acceptedAnswer":{"@type":"Answer","text":"SimReady Central Library is a free, ungated collection of more than 2,000 simulation-ready OpenUSD assets for robotics work. Reusing these assets alongside newly generated objects reduces the asset preparation needed to assemble environments like an obstacle course in Isaac Sim. Developers assembling simulation scenes can follow asset-library resources like this on daily.dev."}}]} ```