# AI Agents Create Virtual Playgrounds for Robot Training MIT and Toyota's SceneSmith system enables robots to train in realistic virtual environments. - **Source:** [MIT News — AI](https://news.mit.edu/topic/artificial-intelligence2) - **Published:** 2026-07-13 - **Topic:** [Data & AI](https://brifing.si/c/data-ai) - **Original article:** - **On Brifing:** https://brifing.si/s/95e6515f-6d52-4eab-9aa5-2c4bc9dc44ef ## The Data Challenge in Robot Training Robots need extensive and diverse data to be effective in the real world. However, physically collecting this data is time-consuming and costly. ## The Role of Virtual Environments MIT and Toyota's SceneSmith system allows robots to train in realistic virtual environments. This reduces the cost and time of physical testing. ## Vision-Language Model (VLM) SceneSmith uses advanced vision-language models like GPT-5.2 to create virtual environments. These models generate realistic scenes based on text and image data. ## Collaboration of Agents The system features three key agents: designer, critic, and orchestrator. These agents collaborate to create and refine realistic scenes. ## Robots' Virtual Experiences In SceneSmith's virtual environments, robots can practice tasks like carrying cups or placing fruits. This reduces errors in the real world. ## Realism Tests Researchers tested the realism of virtual environments using pre-trained robot policies. Robots behaved in virtual environments as they would in the real world. ## The Future of Robot Training Systems like SceneSmith can revolutionize robot training by making it faster and more efficient. This marks the beginning of a new era in robotics. ## Sources covering this story - https://news.mit.edu/2026/ai-agents-create-virtual-playgrounds-to-help-robots-get-crucial-training-data-0713 — --- _Summary written by Brifing from the original reporting. Copyright in the original article belongs to its publisher; read the full piece at the link above._