Multi-Agent Topology Optimization for Space Software Architectures Using Genetic Algorithms

The efficiency of a multi-agent software architecture fundamentally depends on the organization of its agents and the modeling strategies used to capture their interactions. Achieving rapid and robust consensus among agents serves as a key indicator of successful algorithmic implementation. Furthermore, considerations such as communication reliability and cost impose practical constraints on the solution space, shaping the topological optimization of multi-agent systems. Metaheuristic approaches, renowned for their ease of implementation and flexibility, provide a powerful avenue for addressing these challenges, particularly when combined to surpass the performance of single-algorithm strategies. In this work, we explore the application of randomized optimization techniques, specifically genetic algorithms, to identify optimal interaction topologies in multi-agent architectures, with a focus on satellite software design. The proposed algorithm improves the time to find an optimal solution by at least 80% compared to the bruteforce algorithm's best-case scenario, with a communication cost reduction of 25-50% for the proposed rover navigation scenarios.