首页 > AI前沿 > PCGRL for Complex Platformers: Synthesizing Playable and Diverse Levels via Inflated Envelopes and Forward Curricula

PCGRL for Complex Platformers: Synthesizing Playable and Diverse Levels via Inflated Envelopes and Forward Curricula

arXiv机器学习 2026-11-17 13:55 7 阅读 查看原文
用户需要我把一段英文新闻正文排版成语义化的HTML片段。让我仔细分析要求: 1. 只输出HTML片段本身,不要Markdown围栏、不要解释 2. 每个自然段落用

包裹,长文用

/

划分小节 3. 列表内容用
  • ,引用用
    ,代码用
    ,重要句子可用
    4. 保留原文所有信息,不得增删改内容
    5. 不要输出//等文档级标签
    
    现在看正文内容,这是一篇关于PCGRL(强化学习程序化内容生成)的技术新闻/论文摘要。内容结构:
    - 第一段:介绍PCGRL的背景和问题(探索瓶颈、扩展限制、传统方法在大画布上无法保持关卡可解性)
    - 第二段:提出解决方案框架,包含三个要点("First", "Second", "Finally")
    - 第三段:实验结果
    
    我需要合理划分小节。原文没有明确的标题,但为了语义化,可以添加合适的标题。不过规则说"保留原文所有信息,不得增删改内容",添加标题算不算增?标题是结构性的,不是内容信息,我认为可以添加

    /

    来划分小节,这是规则要求的"长文用

    /

    划分小节"。 让我重新考虑。原文是一整段连贯的技术摘要。可以这样划分: 第一段(问题与背景)可以作为一个小节,加个标题比如"研究背景与问题"——但这样算增内容吗?标题本身是新加的。不过规则明确说"长文用

    /

    划分小节",这意味着需要添加标题。 实际上,我可以把原文的三个逻辑部分分别处理: 1. 背景与问题:从开头到"...on large-scale canvases." 2. 方法框架:从"To address these challenges"到"...rejection filtering mechanism." 3. 实验结果:从"Experimental results"到结尾 对于方法部分,原文有"First", "Second", "Finally"这样的结构,可以用列表或段落来呈现。但原文是完整的段落句子,不是列表项。规则说"保留原文所有信息,不得增删改内容",所以不能把句子拆成列表项而改变结构?或者可以将三个要点用
    • 来呈现,因为它们是并列的要点。 让我再想想。原文:"First, we construct... Second, to alleviate... Finally, to resolve..." 这三个是并列的方法贡献。用
      • 呈现是合理的语义化处理,内容没有改变。 但要注意,每个li里是完整的句子,我需要保持原文完整。 方案:

        研究背景与挑战

        Procedural Content Generation via Reinforcement Learning (PCGRL) has emerged...on large-scale canvases.

        框架设计

        To address these challenges, we present a physics-aware, curriculum-guided PCGRL framework.

        • First, we construct an "Inflated Volume Envelope" mechanism...
        • Second, to alleviate the sparse reward challenge...
        • Finally, to resolve the issue where deterministic policies...
        但这样"To address these challenges, we present..."这个句子被单独提出来了,然后下面用列表。原文是连续的一段话。如果拆分,内容没有丢失,但结构变了。规则允许用
        • 处理列表内容,这三个要点确实符合列表特征。 不过为了更忠实于原文,也许应该保持段落完整。但规则说"列表内容用
          • ",而这三个"First/Second/Finally"确实是列表性质的内容。 我觉得用列表更语义化,也符合规则。内容不增不减,只是用HTML语义化标签表达结构。 关于"重要句子可用",我可以加一些强调,比如框架的三个核心机制名称。但要小心不要过度。可以对"Inflated Volume Envelope"或者关键结论加。 让我写出完整的HTML:

            研究背景与挑战

            Procedural Content Generation via Reinforcement Learning (PCGRL) has emerged as a promising paradigm for automated game development. However, existing PCGRL frameworks still suffer from severe exploration bottlenecks and scaling limitations when faced with long-sequence generation tasks governed by strict rigid-body rules and non-linear kinematic boundaries (e.g., platformer constraints). Consequently, traditional approaches frequently fail to maintain reliable level solvability on large-scale canvases.

            方法框架

            To address these challenges, we present a physics-aware, curriculum-guided PCGRL framework.

            • First, we construct an "Inflated Volume Envelope" mechanism, which provides a physics-constrained representation that incorporates continuous kinematic feasibility into the discrete state space of the MDP.
            • Second, to alleviate the sparse reward challenge induced by extremely long decision sequences, we introduce intermediate state-transition feedback and a multi-stage forward extension curriculum learning framework, enabling informative value estimation and stable credit assignment over long-horizon trajectories.
            • Finally, to resolve the issue where deterministic policies after convergence generate highly monotonous map paths with a lack of diversity, we activate high-entropy stochastic inference during the synthesis phase, regularized by a post-hoc parallel rejection filtering mechanism.

            实验结果

            Experimental results demonstrate that our framework significantly enhances structural diversity while maintaining high physical playability across multi-scale environments. Comparative evaluations across multi-scale canvases demonstrate that our framework maintains superior convergence stability and dimensional scalability, reliably synthesizing solvable levels where traditional PCGRL baselines degrade.

            检查内容是否完整: - 所有原文句子都在 - 没有增删内容 - 标题是结构性的,不算内容增减 - 用

            划分了小节 - 列表用了
            • - 重要句子用了 - 没有文档级标签 这样应该符合要求。让我再检查一下原文,确保一字不差。 原文第一段:Procedural Content Generation via Reinforcement Learning (PCGRL) has emerged as a promising paradigm for automated game development. However, existing PCGRL frameworks still suffer from severe exploration bottlenecks and scaling limitations when faced with long-sequence generation tasks governed by strict rigid-body rules and non-linear kinematic boundaries (e.g., platformer constraints). Consequently, traditional approaches frequently fail to maintain reliable level solvability on large-scale canvases. ✓ 完整 原文第二段开头:To address these challenges, we present a physics-aware, curriculum-guided PCGRL framework. ✓ First部分:First, we construct an "Inflated Volume Envelope" mechanism, which provides a physics-constrained representation that incorporates continuous kinematic feasibility into the discrete state space of the MDP. ✓ Second部分:Second, to alleviate the sparse reward challenge induced by extremely long decision sequences, we introduce intermediate state-transition feedback and a multi-stage forward extension curriculum learning framework, enabling informative value estimation and stable credit assignment over long-horizon trajectories. ✓ Finally部分:Finally, to resolve the issue where deterministic policies after convergence generate highly monotonous map paths with a lack of diversity, we activate high-entropy stochastic inference during the synthesis phase, regularized by a post-hoc parallel rejection filtering mechanism. ✓ 实验结果:Experimental results demonstrate that our framework significantly enhances structural diversity while maintaining high physical playability across multi-scale environments. Comparative evaluations across multi-scale canvases demonstrate that our framework maintains superior convergence stability and dimensional scalability, reliably synthesizing solvable levels where traditional PCGRL baselines degrade. ✓ 全部完整。 现在关于的使用,我对First/Second/Finally加了,对关键结论加了。这符合"重要句子可用"。 标题用中文还是英文?原文是英文,标题用中文可能有点不一致,但规则没有明确要求。为了语义清晰,用中文小节标题是合理的。或者也可以不加标题,直接用段落。但规则说"长文用

              /

              划分小节",所以应该加。 我觉得这样排版是合理的。输出时不加任何Markdown