Affordance-Graphed Task Worlds: Self-Evolving Task Generation for Scalable Embodied Learning

TL;DR

This paper introduces AGT-World, a framework for autonomous, scalable robotic learning that constructs structured task environments and refines policies through self-evolution, improving success and generalization.

Contribution

The paper presents a novel structured graph-based task space and a self-evolution mechanism combining vision-language reasoning and geometric verification for autonomous policy refinement.

Findings

Outperforms existing methods in success rates.

Achieves better generalization in complex tasks.

Demonstrates effective self-improving cycle for robot learning.

Abstract

Training robotic policies directly in the real world is expensive and unscalable. Although generative simulation enables large-scale data synthesis, current approaches often fail to generate logically coherent long-horizon tasks and struggle with dynamic physical uncertainties due to open-loop execution. To address these challenges, we propose Affordance-Graphed Task Worlds (AGT-World), a unified framework that autonomously constructs interactive simulated environments and corresponding robot task policies based on real-world observations. Unlike methods relying on random proposals or static replication, AGT-World formalizes the task space as a structured graph, enabling the precise, hierarchical decomposition of complex goals into theoretically grounded atomic primitives. Furthermore, we introduce a Self-Evolution mechanism with hybrid feedback to autonomously refine policies, combining Vision-Language Model reasoning and geometric verification. Extensive experiments demonstrate that our method significantly outperforms in success rates and generalization, achieving a self-improving cycle of proposal, execution, and correction for scalable robot learning.