AI 中文总结
本文综述2024-2025年软件设计与架构研究，采用主题综合法分析五大领域，指出现代架构需覆盖全生命周期，AI辅助等技术可提升软件特性，同时存在实证验证不足等缺口。
AI 中文摘要
软件架构已随现代软件系统（尤其是基于云计算、微服务、人工智能（AI）和分布式计算环境的系统）日益增长的复杂性发生了显著演变。本文献综述综合了2024至2025年间发表的最新研究，以考察软件设计与架构领域的新兴趋势、挑战及未来方向。该综述采用主题综合方法，对五个主要领域的当代研究进行分析：架构建模与表示、软件质量属性与自适应架构、架构演进与复杂性管理、人工智能辅助架构决策，以及现有研究缺口。研究结果表明，现代软件架构已超越传统结构设计范畴，需在整个软件生命周期中支持持续架构治理、利益相关者沟通、运行时可观测性、弹性及智能决策支持。此外，所综述的研究显示，多种架构视图、持续监控、领域驱动分解及AI辅助设计技术，对提升可扩展性、可维护性、适应性及长期软件可持续性具有重要作用。尽管取得了这些进展，仍存在若干研究缺口，包括所提方法的实证验证有限、安全与隐私在架构决策中的整合不足、对边缘与无服务器计算等新兴范式的探索不足，以及缺乏用于可信AI辅助架构的标准化框架。
英文摘要
Software architecture has evolved considerably in response to the increasing complexity of modern software systems, particularly those based on cloud computing, microservices, artificial intelligence (AI), and distributed computing environments. This literature review synthesizes recent studies published between 2024 and 2025 to examine emerging trends, challenges, and future directions in software design and architecture. The review adopts a thematic synthesis approach to analyse contemporary research across five major areas: architectural modelling and representation, software quality attributes and self-adaptive architectures, architectural evolution and complexity management, artificial intelligence-assisted architectural decision-making, and existing research gaps. The findings indicate that modern software architecture extends beyond traditional structural design to support continuous architectural governance, stakeholder communication, runtime observability, resilience, and intelligent decision support throughout the software lifecycle. Furthermore, the reviewed studies demonstrate that multiple architectural views, continuous monitoring, domain-driven decomposition, and AI-assisted design techniques contribute significantly to improving scalability, maintainability, adaptability, and long-term software sustainability. Despite these advances, several research gaps remain, including limited empirical validation of proposed approaches, insufficient integration of security and privacy into architectural decision-making, inadequate exploration of emerging paradigms such as edge and serverless computing, and the absence of standardized frameworks for trustworthy AI-assisted architecture.
