Two extensions of heterogeneous boid model to avoid metastable patterns
- PDF / 2,012,306 Bytes
- 10 Pages / 595.276 x 790.866 pts Page_size
- 2 Downloads / 159 Views
ORIGINAL ARTICLE
Two extensions of heterogeneous boid model to avoid metastable patterns Mari Nakamura1 Received: 20 May 2020 / Accepted: 13 September 2020 / Published online: 19 October 2020 © International Society of Artificial Life and Robotics (ISAROB) 2020
Abstract A boid is a simple multiagent model of animal group behavior. Boid agents communicate locally. I studied a heterogeneous boid model comprised of many agents that are divided into several types. While varying interaction among types of agents, this model generates stable patterns with a symmetric interface among different types of agents. As the number of agents increases and as agent clusters grow larger, this model forms a metastable pattern (i.e., a complex of such stable patterns) that is caused by conflict among the local growths of stable patterns. To avoid metastable patterns, I designed two extended heterogeneous boid models: a two-component boid with noise control and a three-component boid with a type transition of agents. In this paper, I examined how these extended models rearrange agents from metastable patterns into stable patterns. These extended models generate stable patterns regardless of the agent number in a shorter time than the original heterogeneous boid model. Keywords Heterogeneous boid · Symmetry · Scalability · Stability
1 Introduction In a previous paper I extended a heterogeneous multiagent model based on boids to generate scalable stable patterns regardless of the number of agents [1]. In this paper, I explained the mechanisms of two extended models and evaluated their resilience. The extended models recovered the stable patterns in a shorter time than the original did. A multiagent system is a tool for studying synergy among many agents. There are two approaches to achieve a multiagent system that works functionally. – The first approach is the cooperation of agents controlled individually. With an increase of agents, calculation for task allocation and communication among agents explode. This work was presented in part at the 3rd International Symposium on Swarm Behavior and Bio-Inspired Robotics (Okinawa, Japan, November 20-22, 2019). * Mari Nakamura tagami‑[email protected] 1
Biomedical Research Institute, National Institute of Advanced Industry and Technology (AIST), Osaka, Japan
13
Vol:.(1234567890)
– The second approach is the control of the group of replaceable agents. This limits communication among agents and fine-task allocation. I improved the second approach by escalating the heterogeneity among agents. The purpose of my study is to develop a self-organizing heterogenous multiagent system. I studied heterogeneous multiagent systems in which the agents are divided into several types. Agents of the same type are homogeneous and replaceable, although not agents of different types. Agents of one type act on the given rule for them and execute the task assigned to them. The rule forms a mass of each type of agents, and decides an action for the agent mass. I constructed behavior rules for heterogeneous
Data Loading...
