[论文] SpecMAGIC: Specification-Guided Multi-Task Generalization in Multi-Age...

论文概要

研究领域: ML
作者: Zijian Guo, İlker Işık, H. M. Sabbir Ahmad, Wenchao Li
发布时间: 2026-10-24
arXiv: 2610.30443

中文摘要

多任务泛化是多智能体强化学习（MARL）中重要却少有人研究的问题：许多真实应用要求多智能体系统泛化到具有多样智能体间协调需求的复杂、时序延展任务。本文提出一种规约语言，把时序逻辑任务规约与表达此类需求的算子结合；并提出 SpecMAGIC 分层框架，利用任务规约的自动机结构做子任务分解——集中式高层规划器识别并跨智能体分配兼容的子任务序列，去中心化低层策略执行单个子任务。这使得系统无需重训练、仅按新规约组合可复用的子任务策略，即可零样本泛化到新任务。文中证明了高层规划器的可靠性与概率完备性。多智能体导航实验表明，SpecMAGIC 在未见过的任务规约上大幅超越现有方法，并随智能体数量增加保持强劲性能。

原文摘要

Multi-task generalization is an important yet underexplored problem in multi-agent reinforcement learning (MARL). Many real-world applications require multi-agent systems to generalize across complex, temporally extended tasks with diverse inter-agent coordination requirements. We introduce a specification language that combines temporal-logic task specifications with operators for expressing such requirements. We propose SpecMAGIC, a hierarchical framework that leverages the automaton structure of task specifications for subtask decomposition. A centralized high-level planner identifies and allocates compatible subtask sequences across agents, while decentralized low-level policies execute individual subtasks. This enables zero-shot generalization to new tasks by composing reusable subtask policies according to new specifications, without retraining. We establish soundness and probabilistic completeness of the high-level planner. Experiments on multi-agent navigation tasks show that SpecMAGIC substantially outperforms existing approaches on unseen task specifications and maintains strong performance as the number of agents increases.

自动采集于 2026-10-24

#论文 #arXiv #ML #小凯