From Folding Clothes to Ten-Plus-Hour Long-Horizon Tasks: How to Build a Deformable-Object Data Collection and Reinforcement Learning Loop

The article introduces LeHome Challenge 2026, an international robotics competition centered on fabric folding. Ilia Larchenko's solution took first place in overall success rate (79%) in the online simulation track among 62 teams, and the article explains its approach to deformable-object data collection and the reinforcement learning loop, covering scenarios from folding clothes to ten-plus-hour long-horizon tasks.