A modular multi-agent Large Language Model pipeline that automates key steps of the systems engineering lifecycle - from requirement structuring and compliance checking to code and test generation - using specialized LLM agents orchestrated within a unified architecture.
Abstract
The increasing complexity of modern software-intensive systems, particularly in the automotive domain, demands new approaches to bridge the gap between high-level engineering specifications and executable, safety-compliant code. This need is amplified by the rapid transition toward software-defined vehicles, where highly dynamic, updateable software functions significantly enlarge the scope and frequency of engineering activities and require scalable, transparent, and adaptive development processes. While recent advances in Large Language Models have demonstrated strong capabilities in automating tasks such as requirements analysis, code generation, and documentation, their deployment in safety-critical engineering workflows remains challenging due to the need for transparency, traceability, and controlled decision-making.This paper presents a modular multi-agent Large Language Model (LLM) pipeline that automates key steps of the systems engineering lifecycle - from requirement structuring and compliance checking to code and test generation - using specialized LLM agents orchestrated within a unified architecture. A central contribution of this work is the integration of a Human-in-the-Loop subsystem, which introduces configurable review checkpoints at critical stages such as requirements analysis, compliance assessment, code generation, and test creation. The human-in-the-loop module enables engineers to approve, reject, or modify intermediate results, ensuring human oversight, enhancing trustworthiness, and enabling adherence to functional safety standards.The system supports heterogeneous input formats and provides end-to-end traceability through structured outputs and detailed monitoring of performance metrics including model usage, token consumption, and automation efficiency. Initial evaluations indicate that the combination of multi-agent specialization and human-in-the-loop-guided oversight can significantly reduce engineering effort while maintaining the transparency and reliability required for regulated domains.By embedding controllable human supervision into the LLM-driven pipeline, this work offers a practical and scalable architecture for integrating Artificial Intelligence (AI) automation into safety-critical systems engineering processes, with particular relevance to automotive software development.
Software complexity is a long-standing challenge for system engineers. Model-Driven Engineering (MDE) addresses it by treating models as first-class artefacts, but a typical MDE process spans many tools and produces heterogeneous models of different system aspects, making traceability, maintenance, and change management difficult. We propose RADIANT, an engineering methodology that combines MDE with Multi-Agent Large Language Models (LLMs) for complete model-based system development, with a focus on safety-critical systems. From a carefully specified requirement model, RADIANT automatically generates heterogeneous models across engineering phases -- a concept model, a domain-specific modelling language, a conforming system model, and a behaviour model -- together with executable, element-level traceability links, on top of which it provides exact, automated change-impact analysis. Generated behaviour models are translated into CSP and formally verified (e.g.\ for deadlock freedom and convergence) with a counterexample-driven repair loop. Evaluating RADIANT across three LLMs, we find that the multi-agent decomposition reliably improves the \emph{syntactic validity} of the generated formal artefacts over a single-agent baseline -- and their \emph{executability} where the model's code generation permits -- while gains in semantic accuracy are model-dependent. A six-participant study shows an order-of-magnitude ($10$--$15\times$) reduction in development time, and the unmodified pipeline transfers to a second domain.
Ran Wei, Letian Zhu, Haochi Wang et al.· 0 citations
The automotive industry is experiencing an unprecedented growth in software complexity driven by electrification, autonomy, and connectivity. Model-Based Design (MBD) using graphical development tools such as Simulink from MathWorks has become standard in the industry, but the translation of natural language software requirements into compliant, tested control models remains a largely manual and time-consuming task. This paper presents a multi-agent large language model (LLM) framework tailored for automotive model-based software development. Our framework orchestrates a set of specialized AI agents to cover the implementation and model-level verification phases of the software development lifecycle, including Simulink model generation, test case creation, compliance checking, test case execution, and automated refinement. To rigorously assess our approach, we introduce the Automotive Software Engineering Benchmark (ASWE-Bench), a new benchmark suite with 38 automotive software requirements categorized into data transformations, combinational logic, stateful logic with timers, and closed-loop control. Our results demonstrate a substantial improvement in generative AI performance: the rulebook-equipped single-agent baseline achieve only a 47.4% model-level pass rate, while our full multi-agent framework attains 73.7% across generated models. This work advances generative AI from isolated prompting to automation in complex development environments and sets a foundation for scalable AI-driven software pipelines in model-based systems engineering. To support reproducibility and future benchmarking, the proposed ASWE-Bench, including requirements and unit tests, is publicly available at: https://git-ce.rwth-aachen.de/mmp-rwth-aachen/aswe-bench
Abdelrahman Abdalla, Vincent Thie, Joschka Schaub et al.· IEEE Access· 1 citation
Creating code specifications is a crucial measure to improve the trustworthiness of many industrial systems implemented in Rust with high security requirements. Because writing specifications requires highly specialized professionals and is time-consuming, the automatic generation of specifications, enabled by large language models (LLMs), has received increasing attention and shown promising results. However, these methods typically focus on partial specification generation (generating proofs after the contract is known) and on extracting dependencies between code modules using predefined relations. This is not suitable for real-world industrial systems where the goal is to generate complete specifications from scratch and where the complex dependencies between code modules are variable. To address this, we propose a multi-agent collaborative framework, StarVerus, to automate the verification of industrial Rust code. Specifically, StarVerus addresses the aforementioned limitations in two ways: 1) In the generation phase, it instructs the LLM to generate all specifications for a given code, and in the repair phase, it uses a cascaded two-stage process of contract alignment and proof repair to correct them; 2) In both the generation and repair phases, it utilizes a function call graph to adaptively obtain bidirectional contextual information (i.e., what it calls and what calls it) for each code module as an additional information source for the LLM. Furthermore, StarVerus introduces a planner-repairer-actor-rewriter multi-agent paradigm to further enhance the proof repair capabilities. Finally, the effectiveness of StarVerus is validated through experiments on benchmark datasets and deployment in a real operating system.
Chao Jiang, Ding Wang, Dugang Liu et al.· Proceedings of the 32nd ACM...· 0 citations
A comprehensive overview of the existing tools and frameworks for implementing MAS in software engineering and a set of lessons learned and challenges that can help researchers and practitioners to select a suitable MAS framework according to their needs are provided.
Mariama Celi Serafim De Oliveira, M. Ibiyo, Marco Gianrusso et al.· 0 citations
In modern software development, the rapid advancement of Large Language Models (LLMs) has made the end-to-end transformation of Natural Language Requirements (NLRs) into executable repository-level code increasingly feasible. However, existing approaches typically rely on simplified instructions (e.g., single-sentence descriptions), failing to reflect complex software development scenarios. Moreover, they lack explicit requirement traceability mechanisms, making it difficult to precisely align and validate generated code against original requirements. To address these limitations, we propose TraceDev, a multi-agent framework for automated software development grounded in use cases that contain multiple functional points and complex semantics. TraceDev employs five role-specific agents, including a Requirement Refiner, Designer, Developer, Tester, and Validator. Notably, the Validator Agent constructs and maintains a heterogeneous traceability graph that links requirements, design models, and code artifacts for interacting with the preceding four agents. The traceability graph maintains consistency across various artifacts and serves as a structured context for efficient memory management, supporting reliable repository-level code generation. We evaluate TraceDev on two widely used datasets (including 125 use cases) compared with two state-of-the-art approaches. On the ETOUR dataset, TraceDev achieves a success rate of 53.63\%, outperforming baseline approaches by up to 186.63\%. A similar trend is observed on the SMOS dataset, where TraceDev attains a success rate of 56.82\%, exceeding baseline approaches by up to 340.80\%. These results demonstrate the effectiveness of TraceDev in repository-level code generation from requirements.
Mingyu Chen, Yakun Zhang, Zihao Xie et al.· 0 citations
The evolution from reactive to proactive AI systems represents a paradigm shift in software engineering, enabling autonomous agents to anticipate requirements, plan complex workflows, and execute multi-step development tasks without human intervention. This paper presents a novel multi-agent architecture for autonomous code generation and software maintenance in enterprise environments. Our framework integrates specialized AI agents for requirements analysis, code synthesis, testing, and deployment, coordinated through a hierarchical planning system with adaptive decision-making capabilities. The system employs reinforcement learning to optimize task allocation, learn from code review feedback, and improve over time. Experimental evaluation on real-world enterprise codebases demonstrates that our proactive agent system achieves 87.3% automated issue resolution, reduces bug fix latency by 62%, and maintains code quality metrics comparable to human developers. The framework successfully handles complex refactoring tasks, security vulnerability remediation, and feature implementation with minimal human oversight, representing a significant advancement toward fully autonomous software engineering workflows.
Saher Elsayed, Samer Abubaker, M. Ali et al.· Annual International Compute...· 0 citations