Workshop Overview

This workshop explores the use of synthetic data as a catalyst for robotic learning, enabling progress across reinforcement learning, system identification, evaluation, and reasoning. By bringing together methodologies from control, machine learning, and computer graphics, the workshop highlights how synthetic data can accelerate development, improve robustness, and expand the versatility of embodied AI systems. Discussions will address both the opportunities and challenges of leveraging synthetic data, including domain transfer, realism tradeoffs, and alignment with real-world deployment. The scope spans a broad spectrum of applications, from manipulation to legged robots, with invited speakers offering diverse perspectives on how synthetic data can shape the future of reliable, adaptable, and trustworthy robotic intelligence. 

Call for Extended Abstract

We invite submissions on topics related to synthetic data for robot learning, including but not limited to:

• Physics-based simulation and synthetic data generation

• Sim-to-real transfer and domain adaptation

• Reinforcement learning with synthetic data

• System identification and calibration

• Evaluation and benchmarking using synthetic environments

• Deformable and rigid body simulation for robotics

• Synthetic data for perception, manipulation, and locomotion

We welcome the abstract submissions and present by posters:

• Extended abstracts (2-4 pages): Work in progress or previously published work