Learning From Insects to Increase Multi-Agent System Resilience: Functional Decomposition and Transfer to Support Biologically Inspired Design

Abstract

Resilience is an emergent property of complex systems that describes the ability to detect, respond, and recover from adversity. Much of the modern world consists of multiple, interacting, and independent agents (i.e. Multi-Agent Systems). However, the process of improving Multi-Agent System resilience is not well understood. We seek to address this gap by applying Biologically Inspired Design to increase complex system resilience. Eusocial insect colonies are an ideal case study for system resilience. Although individual insects have low computing power, the colonies collectively perform complex tasks and demonstrate resilience. Therefore, analyzing key elements of eusocial insect colonies may offer insight on how to increase Multi-Agent System resilience. Before the strategies used in eusocial insects can be adapted for Multi-Agent Systems, however, the existing research must be identified and transferred from the biological sciences to the engineering field. These transfers often hinder or limit biologically inspired design. This paper translates the biological investigation of individual insects and colony network behavior into strategies that can be tested to increase Multi-Agent System resilience. These strategies are formulated to be applied to Agent-Based Modeling. Agent-Based Modeling has been applied to many Multi-Agent Systems including epidemiology, traffic management, and marketing. This provides a key step in the design-by-analogy process: Identifying and decoding analogies from their original context. The design principles proposed in this work provide a foundation for future testing and eventual implementation into Multi-Agent Systems.