Swarm Intelligence

From Social Bacteria to Humans

Andrew Schumann

• Éditeur: CRC Press
• Année: 2020
• ISBN:
  9780429647604
• Édition: 1
• Sujet: Ciencia de la computación
• Sous-sujet: Minería de datos

The notion of swarm intelligence was introduced for describing decentralized and self-organized behaviors of groups of animals. Then this idea was extrapolated to design groups of robots which interact locally to cumulate a collective reaction. Some natural examples of swarms are as follows: ant colonies, bee colonies, fish schooling, bird flocking, horse herding, bacterial colonies, multinucleated giant amoebae Physarum polycephalum, etc. In all these examples, individual agents behave locally with an emergence of their common effect.

An intelligent behavior of swarm individuals is explained by the following biological reactions to attractants and repellents. Attractants are biologically active things, such as food pieces or sex pheromones, which attract individuals of swarm. Repellents are biologically active things, such as predators, which repel individuals of swarm. As a consequence, attractants and repellents stimulate the directed movement of swarms towards and away from the stimulus, respectively.

It is worth noting that a group of people, such as pedestrians, follow some swarm patterns of flocking or schooling. For instance, humans prefer to avoid a person considered by them as a possible predator and if a substantial part of the group in the situation of escape panic (not less than 5%) changes the direction, then the rest follows the new direction, too. Some swarm patterns are observed among human beings under the conditions of their addictive behavior such as the behavior of alcoholics or gamers.

The methodological framework of studying swarm intelligence is represented by unconventional computing, robotics, and cognitive science. In this book we aim to analyze new methodologies involved in studying swarm intelligence. We are going to bring together computer scientists and cognitive scientists dealing with swarm patterns from social bacteria to human beings. This book considers different models of simulating, controlling, and predicting the swarm behavior of different species from social bacteria to humans.

1

Introduction

Andrew Schumann

Department of Cognitive Science and Mathematical Modelling, University of Information Technology and Management in Rzeszow, Sucharskiego 2, 35-225 Rzeszow, Poland

Email: [email protected]

The notion of swarm intelligence [6, 41] was introduced to describe the decentralized and self-organized behaviors of groups of animals. This idea was then extrapolated to design groups of robots which interacted locally to cumulate a collective reaction. Some natural examples of swarms are as follows [38]: ant colonies, bee colonies, fish schooling, bird flocking, horse herding, bacterial colonies, multinucleated giant amoebae Physarum polycephalum, etc. In all these examples, individual agents behave locally with an emergence of their common effect.

At first, swarm intelligence was studied in order to develop new algorithms in transporting and scheduling – the point being that ants, bees, some social bacteria, Physarum polycephalum, etc. can solve logistic problems very effectively [38]: (i) the Travelling Salesman Problem can be solved by ants and by amoebae; (ii) the Steiner Tree Problem can be solved by amoebae; (iii) the Generalized Assignment Problem can be solved by bees; (iv) mazes can be solved by ants and by amoebae, etc. intelligent behavior of swarm individuals is explained by the following biological reactions to attractants and repellents [5, 14, 30, 35]. Attractants are biologically active things, such as food pieces or sex pheromones, which attract individuals of the swarm. Repellents are biologically active things, such as predators, which repel individuals of the swarm. As a consequence, attractants and repellents stimulate the directed movement of swarms towards and away from the stimulus, respectively.

It is worth noting that a group of people, such as pedestrians, follow some swarm patterns such as flocking or schooling. For instance, humans prefer to avoid a person considered by them as a possible predator and if a substantial part of the group in the situation of escape panic (not less than 5%) it changes the direction, then the rest of the group follows the new direction, too. Some swarm patterns are observed among human beings under the conditions addictive behavior such as the behavior of alcoholics or gamers [38].

The methodological framework of studying swarm intelligence is represented by unconventional computing, robotics, and cognitive science. In this book we aim to analyze new methodologies involved in studying swarm intelligence. We are going to bring together computer scientists and cognitive scientists dealing with swarm patterns from social bacteria to human beings.

In modeling swarms, we assume that animal collectives can contain different numbers of their members – from a small number to a large enough number. For example, one cluster of naked mole-rats includes on average from 75 to 80 individuals, while there are ant colonies consisting of many million worker ants and many thousand queen ants living in many thousand nests. The task of simulating multi-agent systems with many millions of actors evidently is quite hard. But standard networks, we deal with in our life, such as social networks contain so many active individuals, too.

A neuromorphic computer, to which we have devoted the first contribution to this book entitled Swarm Intelligence for Morphogenetic Engineering, can be represented as a swarm with a huge number of active components. For designing this computer we should set 100 billion neurons and define 100 trillion nonrandom connections among them. These neurons are regarded by Bruce J. MacLennan and Allen C. McBride (the two authors of the chapter) as separate microscopic agents (microrobots) that can emit and respond to simple signals and implement simple control processes, but they can also move and transport other components. Signaling molecules and structural components are considered passive components, because they cannot move without external forces. Microrobots as active components take part in an artificial morphogenesis through assembling passive components into a desired structure. This morphogenesis is a result of interactions of collectives of microrobots.

In this chapter, Bruce J. MacLennan and Allen C. McBride proposed certain algorithms for the coordination of microrobot swarms involved in morphogenetic engineering. The morphogenetic programming notation for this purpose is based on a mathematical notation developed for partial differential equations. For more details on neuromorphic computers with artificial morphogenesis, please see [24, 25, 26, 27, 28, 29]. Hence, the ideas of Bruce J. MacLennan and Allen C. McBride appeals to the modeling and controlling of artificial swarms consisting of millions microrobots.

In this chapter, Bruce J. MacLennan and Allen C. McBride proposed certain algorithms for the coordination of microrobot swarms involved in morphogenetic engineering. The morphogenetic programming notation for this purpose is based on a mathematical notation developed for partial differential equations. For more details on neuromorphic computers with artificial morphogenesis, please see [24, 25, 26, 27, 28, 29]. Hence, the ideas of Bruce J. MacLennan and Allen C. McBride appeals to the modeling and controlling of artificial swarms consisting of millions microrobots.

Some computational tasks which are being solved by swarms effectively such as transporting and scheduling are studied in depth. Nevertheless, there are many sophisticated tasks solved by swarms daily which are little known in computer science, yet. For instance, among social insects we can observe necrophoresis – a social phenomenon of carrying dead bodies of members of colonies of ants or bees from the nest [8, 15]. In the second contribution to the book under the title Ant Cemeteries as a Cluster or as an Aggregate Pile prepared by Tomoko Sakiyama, there is an examination of a formation of cemeteries performed by ant workers. This social behavior can be formalized by some simple clustering rules such as the following implication: if ant workers find a corpse, then they pick up it with a probability that decreases according to the cluster size, while corpse-carrying ants drop their carrying corpses with a probability that increases due to the cluster size. On the basis of these rules, ant workers build large piles of corpses [37, 40]. In the model proposed by Tomoko Sakiyama, agents can modify the probability of the drop, which was dependent on whether they detected or did not detect their nest-mates. Therefore, this chapter shows that swarms of ants comply with many from for food to cemeteries.

In a flock of birds and school of fish, individuals try to coordinate their behavior on the basis of their neighbors to avoid collisions with them. However, there are many examples of group behavior without this mechanism. The soldier crabs of Mictyris guinotae behave as a swarm with an internal noise and/or anticipation. In the chapter Robust Swarm of Soldier Crabs, Mictyris guinotae, Based on Mutual Anticipation which was prepared by Y.-P. Gunji, H. Murakami, T. Niizato, Y. Nishiyama, K. Enomoto, A. Adamatzky, M. Toda, T. Moriyama and T. Kawai, there is a kinetic model analyzing how crabs move, revealing dynamic internal structures within groups, such as topological distances, scale-free correlations and inherent noise. To learn more about this model, see [12, 13].

In computer science, there are in general some basic patterns formalized of different swarms: Ant Colony Optimization [9, 10, 11], Artificial Bee Colony [16, 17, 18], Particle Swarm Optimization [19, 20], etc. The algorithms of Ant Colony Optimization and Artificial Bee Colony are aimed, first of all, for solving logistic problems such as scheduling, assignment, and transport. The algorithms of Particle Swarm Optimization allows us to simulate the group movement of animals with collision avoidance (individuals avoid a collision with neighbors), velocity matching (individuals synchronize their speed with their neighbors), and swarm centering (individuals stay close to their neighbors). All these algorithms formalizing the fundamental swarm patterns of animals can be applied in different areas of computer science: from designing artificial swarms of robots to cybersecurity. In the contribution to our book entitled Swarm intelligence in Cybersecurity submitted by Cong Truong Thanh, Quoc Bao Diep, and Ivan Zelinka, there are analyzed prospects of applying swarm intelligence techniques such as Ant Colony Optimization, Particle Swarm Optimization, a...