发表机构
State Key Laboratory of Complex & Critical Software Environment, National University of Defense Technology（复杂与关键软件环境国家重点实验室，国防科技大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
提出R2SGEN框架，通过解耦SMT约束求解和定制剪枝，从真实数据合成物理合理的场景程序，在nuScenes上优于LLM基线。
AI 中文摘要
合成训练数据的需求受到 sim-to-real 差距的阻碍，因为当前基于数据驱动和基于 LLM 的生成器常常产生物理上不合理的场景。为了解决这个问题，我们提出了 R2SGEN，一个 Real-to-Sim 框架，它从真实世界数据中合成结构化的场景程序。为了克服整体式可满足性模理论（SMT）编码的组合爆炸和难以处理的问题，我们引入了一种解耦合成策略。该方法使用轻量级的原子 SMT 约束，将离散的结构程序搜索与连续的几何求解分开。此外，我们通过集成两种定制的剪枝机制显著加速了搜索过程：用于广度优先搜索的公共前缀抽象剪枝和用于深度优先搜索的分支定界。我们在 nuScenes 数据集上对 20 个不同复杂度的真实世界场景评估了 R2SGEN。实验结果表明，我们的方法保证了与输入场景的一致性，并在评估的输入下产生了比基于 LLM 的基线显著更低成本的程序。所提出的两种搜索范式展现出互补优势，证明了对高复杂度合成数据生成的高效性和可扩展性。
英文摘要
The demand for synthetic training data is hindered by the sim-to-real gap, as current data-driven and LLM-based generators often produce physically implausible scenarios. To address this, we propose R2SGEN, a Real-to-Sim framework that synthesizes structured scenario programs from real-world data. To overcome the combinatorial explosion and intractability of monolithic Satisfiability Modulo Theories (SMT) encoding, we introduce a decoupled synthesis strategy. This approach separates the discrete structural program search from continuous geometric resolution using lightweight, atomic SMT constraints. Furthermore, we significantly accelerate the search process by integrating two tailored pruning mechanisms: Common Prefix Abstraction-based pruning for Breadth-First Search and Branch-and-Bound for Depth-First Search. We evaluate R2SGEN on 20 real-world scenes of varying complexity from the nuScenes dataset. Experimental results show that our method guarantees consistency with the input scene and produces substantially lower-cost programs than the LLM-based baselines under the evaluated inputs. Both proposed search paradigms exhibit complementary advantages, proving highly efficient and scalable for high-complexity synthetic data generation.
Comments31 pages, 4 figures, 4 tables. Accepted at OOPSLA 2026, to appear in PACMPL Vol. 10, No. OOPSLA2, Article 395 (DOI: 10.1145/3839527). This is the full version with supplementary material
