An Artificial Honeybee Colony Algorithm to Quantify Adaptability via Resilience for Space System Architectures
Multidisciplinary nature-inspired approaches for designing, understanding, optimizing, and representing complex systems are becoming increasingly common in engineering applications. The highly interconnected relationships within these systems are challenging to model; thus, novel techniques to measure performance and optimize system design and function are needed. To explore agent-based biomimetic models to represent distributed systems such as satellite swarms and constellations, this dissertation advances nature-inspired agent-based modeling approaches by surveying and developing a resilience quantification analysis applicable to multiple satellite constellation models. Resilience is a temporal, emergent capability to robustly absorb and adapt to disruptive events, restoring a system's operations or performance requirements. The metric is quantified as a ratio of the variances of disrupted to ideal performance-time data, where retaining a constant uncertainty indicates reduced noise in adaptivity. As a method for analyzing and optimizing resilience, this work presents a novel biological-inspired optimization algorithm, the adaptive artificial honeybee colony (AHC), to search for and generate resilient (or optimal) solutions to highly interconnected or nonlinear problems. The AHC is an agent-based algorithm that integrates pollination models, particle swarm dynamics, and mutualistic relations to generate new solution spaces and locate optimal or resilient solutions. For verification, the AHC's capabilities were tested against particle swarm optimization and gradient descent for five benchmark functions given three different initial guesses ranging in proximities to the optimal solution. The AHC outperformed the other two methods in all five tests, locating optimal solutions in every case regardless of the initial guess' proximity to the optimum. The AHC's optimization capabilities extend to analyzing the resilience of an investigative satellite swarm by maneuvering the satellites to a more distributed, resilient architecture that preserves observation-based performance quality despite disruptive events or satellite failure. Additionally, a sensitivity analysis determined the most sensitive tunable parameter to be the pollination cluster radius, which determines the area new solutions appear within at each iteration via pollination. The results of these tests and applications demonstrate how the AHC's adaptive characteristics are beneficial in optimizing the resilience of highly interconnected, nonlinear, or complex problems where the user may have little to no former knowledge or intuition.