Harnessing Multi-Agent LLMs for Complex Engineering Problem-Solving: A Framework for Senior Design Projects

TL;DR

The paper addresses solving complex, multidisciplinary senior design projects in engineering education, where technical performance must be balanced with ethical, social, and environmental considerations in a globalized context. It proposes a multi-agent LLM framework (MAS LLM) with specialized, role-based agents representing diverse expert perspectives, coordinated by a central coordinator, to simulate interdisciplinary teamwork and robust problem-solving. Through comparative evaluation against a single-agent TOT-based prompt and with faculty and NLP-based metrics, the MAS approach generally aligns more closely with faculty judgments and yields richer, more cohesive feedback. The findings support the framework’s pedagogical value for SDP coaching and its potential for scalable, interdisciplinary engineering education, while acknowledging limitations and outlining directions for future enhancements.

Abstract

Multi-Agent Large Language Models (LLMs) are gaining significant attention for their ability to harness collective intelligence in complex problem-solving, decision-making, and planning tasks. This aligns with the concept of the wisdom of crowds, where diverse agents contribute collectively to generating effective solutions, making it particularly suitable for educational settings. Senior design projects, also known as capstone or final year projects, are pivotal in engineering education as they integrate theoretical knowledge with practical application, fostering critical thinking, teamwork, and real-world problem-solving skills. In this paper, we explore the use of Multi-Agent LLMs in supporting these senior design projects undertaken by engineering students, which often involve multidisciplinary considerations and conflicting objectives, such as optimizing technical performance while addressing ethical, social, and environmental concerns. We propose a framework where distinct LLM agents represent different expert perspectives, such as problem formulation agents, system complexity agents, societal and ethical agents, or project managers, thus facilitating a holistic problem-solving approach. This implementation leverages standard multi-agent system (MAS) concepts such as coordination, cooperation, and negotiation, incorporating prompt engineering to develop diverse personas for each agent. These agents engage in rich, collaborative dialogues to simulate human engineering teams, guided by principles from swarm AI to efficiently balance individual contributions towards a unified solution. We adapt these techniques to create a collaboration structure for LLM agents, encouraging interdisciplinary reasoning and negotiation similar to real-world senior design projects. To assess the efficacy of this framework, we collected six proposals of engineering and computer science of...

Figures (7)

• Figure 1: Workflow Overview: 1) Input Submission: The student enters the project title and proposal PDF. 2) Proposal Processing: Coordinator Agent forwards details to Tasks Agent to generate focused tasks. 3) Task Generation: Tasks Agent returns a list of tasks. 4) Task Distribution: Tasks are sent to the Tasks Channel. 5) Task Assignment: Tasks are assigned to relevant agents. 6) Output Generation: Agents produce outputs and send them to the Tasks Channel and Coordinator Agent. 7) Input Linking: Outputs from one agent can serve as inputs for others if needed. 8) User Output: Final results are displayed in the interface, with summaries and detailed analysis.
• Figure 2: Personas for each agent in the proposed MAS, detailing their tasks, objectives, and evaluation points for evaluating engineering SDPs. The specialized agents—covering problem formulation, breadth and depth, ambiguity, complexity, innovation, ethics, and methodology—enable systematic and holistic assessments across technical and non-technical dimensions.
• Figure 3: Visual representation of scores to evaluate the performance of both of our proposed multi-agent and single-agent systems across each SDP (represented in numbers on the x-axis) for each aspect of engineering and computing. It can be seen that multi-agent system scores are more aligned with faculty evaluation scores as compared to single-agent scores.
• Figure 4: Mean Absolute Error comparison between multi-agent and single-agent systems against faculty evaluations. Lower bars indicate closer alignment with faculty scores, with the multi-agent system generally showing better accuracy.
• Figure 5: NLP-based performance evaluation of original student proposals, MAS responses, and single-agent system responses. The metrics include Clause Density (complexity), Lexical Cohesion (thematic unity), Flesch-Kincaid Score (readability), and Average Sentence Length (structural depth), designed to balance accessibility with scholarly depth. Our results show that the MAS approach consistently outperforms the single-agent system across all evaluated metrics.