Date of Award

2026

Document Type

Open Access Dissertation

Degree Name

Doctor of Philosophy in Mechanical Engineering-Engineering Mechanics (PhD)

Administrative Home Department

Department of Mechanical and Aerospace Engineering

Advisor 1

Jung Yun Bae

Committee Member 1

Myoungkuk Park

Committee Member 2

Vinh T. Nguyen

Committee Member 3

Ashraf Saleem

Abstract

This dissertation focuses on developing coordination strategies for collaborative heterogeneous Multi-Robot Systems (MRS) operating under realistic operational constraints. This research addresses a series of increasingly complex task allocation and path-planning problems that incorporate robot heterogeneity and practical constraints, including size limitations, fuel capacity, and payload restrictions.

Initially, robot-specific size constraints have been introduced to the MRS coordination problem to model accessibility limitations in narrow field passages. Three heuristic approaches are presented to address the problem: a primal-dual workload-balancing method, a greedy heuristic with iterative refinement, and a Large Language Model (LLM)-based heuristic using structured prompt engineering. Results demonstrate that the greedy heuristic achieves superior solution quality, while the LLM-based approach exhibits flexibility and adaptability.

The framework is then extended to incorporate fuel constraints. A constraint-aware greedy heuristic is developed to handle heterogeneous robot capabilities subject to multiple simultaneous operational constraints. A structured LLM-based framework is introduced that enables the systematic generation of routing heuristics across different LLMs. Extensive simulations confirm that the greedy heuristic provides robust performance, while LLM-based heuristics achieve competitive solution quality with significantly reduced computation time.

Finally, payload restrictions are integrated into a unified formulation. Four solution approaches are analyzed: a manually derived mixed-integer linear programming (MILP) formulation solved with a commercial optimization solver Gurobi; a solver-friendly MILP formulation generated by Gurobot, an Artificial Intelligence (AI) driven optimization support framework; a constraint-integrated greedy heuristic; and an LLM-based heuristic that encodes all three constraints. Results reveal clear trade-offs: MILP-based methods provide reliable benchmarks for small instances, while heuristic approaches demonstrate strong robustness across all problem scales.

This dissertation provides a comprehensive framework for modeling and solving heterogeneous multi-robot coordination problems. The proposed methodologies highlight the trade-offs between optimality, scalability, and adaptability, and demonstrate the emerging potential of LLMs as flexible heuristic generators.

Available for download on Tuesday, April 13, 2027

Share

COinS