跳转至

PIE-APT:通过增量推理在时序动态知识图谱上进行 abductive 规划

文章背景与核心概要

在开放世界和信息不完备的环境下,在时序动态知识图谱(Temporal Dynamic Knowledge Graphs, TDKGs)上进行规划一直面临着严峻的理论挑战。传统的动作形式化方法经常陷入可判定性问题以及“分支问题”(Ramification Problem),而结构化的溯因推理(abduction)往往需要庞大的组合搜索空间。为了克服这些障碍,本文提出了 PIE-APT 框架。

该框架包含两个核心组件:一是 PIE-Abducer(增量直接推导溯因模块),它通过将目标的逻辑否定注入到一致的描述逻辑(DL)分支中来合成缺失的前提,从而绕过了传统的最小hitting集(MHS)枚举;二是 PIE-APT(溯因规划模块),它采用递归的“生成-测试”(Generate-and-Test)架构,将后向链式 A* 搜索与 PIE-Abducer 相结合,并通过前向链式的时序投射(Temporal Projection)进行验证。

该框架原生运行在表达能力强大的 \(\mathcal{SROIQ}\) 描述逻辑和 OWL 之上,成功保持了可判定性,并在高级 OWL 基准测试中展现出优于经典规划器的性能。


摘要 (Abstract)

Planning over Temporal Dynamic Knowledge Graphs (TDKGs) presents theoretical challenges in open-world environments with incomplete information. Existing action formalisms often face decidability issues and the Ramification Problem, while structural abduction requires expansive combinatorial search spaces.

在开放世界及信息不完备的环境中,在时序动态知识图谱(TDKGs)上进行规划带来了理论上的挑战。现有的动作形式化方法经常面临可判定性问题以及分支问题(Ramification Problem),而结构化溯因推理则需要庞大的组合搜索空间。

We introduce a unified framework with two modules—PIE-Abducer (incremental direct-derivation abduction) and PIE-APT (Abductive Planning for TDKGs)—operating natively on the expressive \(\mathcal{SROIQ}\) Description Logic. Modeling state transitions as non-monotonic updates to deductively closed DL theories, we represent actions natively in OWL. This leverages an incremental reasoner to preserve decidability and natively bypass the Ramification Problem.

我们引入了一个包含两个模块的统一框架——PIE-Abducer(增量直接推导溯因)和 PIE-APT(针对 TDKGs 的溯因规划),它们原生运行在表达能力强的 \(\mathcal{SROIQ}\) 描述逻辑上。我们将状态转换建模为对演绎封闭的 DL 理论的非单调更新,并在 OWL 中原生表示动作。这利用了增量推理机来保持可判定性,并原生绕过了分支问题。

To address incomplete knowledge, PIE-Abducer circumvents Minimal Hitting Set (MHS) enumeration. Instead of combinatorial search, it injects the logical negation of a goal into a consistent DL branch and synthesizes missing premises via direct refutation consequences. PIE-APT employs a recursive Generate-and-Test architecture, interleaving backward-chaining A* search with PIE-Abducer to synthesize both action sequences and abductive assumptions. Candidates undergo strict validation via forward-chaining Temporal Projection to evaluate logical trajectories.

为了解决知识不完备问题,PIE-Abducer 绕过了最小 Hitting 集(MHS)的枚举。它不进行组合搜索,而是将目标的逻辑否定注入到一致的 DL 分支中,并通过直接反驳推论来合成缺失的前提。PIE-APT 采用递归的“生成-测试”架构,将后向链式 A* 搜索与 PIE-Abducer 交织在一起,以合成动作序列和溯因假设。候选方案通过前向链式的时序投射进行严格验证,以评估逻辑轨迹。

We evaluate four OWL benchmarks targeting semantic abilities missing from classical planning: * Parameterized goals with witness search * Mid-search DL entailment * Open-world assumption injection * Adversarial plan synthesis

我们评估了四个针对经典规划所缺少的语义能力的 OWL 基准测试: * 带有见证搜索(witness search)的参数化目标 * 搜索中期的 DL 蕴涵 * 开放世界假设注入 * 对抗性计划合成

Results show qualitative superiority over classical planners and prove our direct-derivation approach significantly outperforms an MHS-faithful baseline in abductive enrichment.

结果表明,该方法在定性上优于经典规划器,并证明了我们的直接推导方法在溯因丰富化方面显著优于符合 MHS 的基线。