Towards LLM-augmented multiagent systems for agile software engineering

Towards LLM-augmented multiagent systems for agile software engineering

연구 분야: Software Development

학회: 2024 39th IEEE/ACM International Conference on Automated Software Engineering (ASE)

A cognitive multi-agent ecosystem designed for efficient software engineering using Agile methodologies can significantly improve software development processes. Key components include the integration of Multi-Agent Systems (MAS) and Large Language Models (LLMs), utilizing Dynamic Context techniques for agent profiling, and Theory of Mind to enhance collaboration. The CogniSim Ecosystem analyzes problems, proposes solutions, constructs and validates plans, and coordinates specialized agents playing roles such as developers, executors, quality checkers, and methodology reviewers. These agents produce documentation, models, and diagrams (e.g., UML) while adhering to predefined quality and performance measures. The ecosystem also simulates the impact of various team configurations on problem-solving effectiveness, helping organizations identify optimal team structures. Case studies and simulations demonstrate its practical applications.CCS CONCEPTS• Computing methodologies → Multi-agent systems; • Software and its engineering → Agile software development.