Simulation of Collective Intelligence of a Multi-Species Artificial Ecosystem Based on Energy Flow

dc.contributor.authorAsgari, Aliakbar
dc.description.abstractCollective intelligence (CI) emerges from local coordination, collaboration and competition among the individuals within a social group. CI mainly results in a global intelligent behavior. One of the fundamental interactional channels within a CI system is energy flow. Each agent within an artificial or physical ecosystem must absorb energy in order to survive, evolve, breed, and reshape its local environment. In addition because the energy resources are limited in the environment, each agent has to compete with other agents to reach the required level of energy. Understanding the internal energy flow can potentially provide a deep insight into internal activities and external emergent behaviors of a given complex system. This study proposes a stochastic scheme for modeling a multi-species prey-predator artificial ecosystem with two levels of food chain. This will enable us to investigate the influence of energy flow on the ecosystem’s lifetime. The proposed model consists of a stationary hosting environment with dynamic weather condition and fruit trees. The inhabitants of this ecosystem are herbivore and carnivore birds each consisting of species. In our model, the collective behavior emerges in terms of flocking with more added rules consist of breeding, competing, resting, hunting, escaping, seeking and foraging behaviors. Using multi-species scheme, we define the ecosystem as a combination of prey and predator species with inter-competition among species within same level of food chain and intra-competition among those belonging to different levels of food chain. Furthermore, in order to model the energy within the ecosystem, some energy variables as functions of behaviors are incorporated in to the model. Finally, a simulation and visualization structure for implementing the proposed model is developed in this study. The experimental results of 11,000 simulations analyzed by Cox univariate analysis and hazard function suggest that only five out of eight behaviors can statistically significant influence the ecosystem’s lifetime. Furthermore, the results of survival analysis show that out of all possible interactions among energy factors, only two of them, interaction between flocking and seeking energies, and interaction between flocking and hunting energies, have statistically significant impact on the system’s lifetime. In addition, software implementation of the proposed framework validates the stability of simulation and visualization architecture. At last regression results using Nelson-Aalen cumulative hazard function and Cox-Snell variable and scaled Schoenfeld residuals test strongly validate our experimental results. To the best of our knowledge, there are three contributions in this research: First, the high level of complexity in the structure of the proposed model in comparison with the other systems which mostly contains only one species of prey, one species of predator and a kind of resource. While this study introduces two species of prey, capability of competition among species, dynamic weather condition with two element of wind and rain and dynamic resources, various behavioral rules such as escaping, breeding, hunting, resting, etc. Energy flow analysis within an artificial ecosystem is the second contribution. To the best of author’s knowledge there is no similar comprehensive model in the previous literature that investigates the life span of a stochastic multi-species predator-prey artificial ecosystem based on energy flow using Survival Analysis method. Lastly, the simulation results show that the flocking and seeking energy and flocking and hunting energy interactions are the most significant interactions which match with the Thompson iii et al. [ 65] observations in the real life. Their findings indicate that in the real life, birds use flocking behavior for better movement, more efficient food searching and social learning. Flocking motion also decrease predation risk as much as the flock size increases.
dc.publisherUniversité d'Ottawa / University of Ottawa
dc.subjectCollective Intelligence
dc.subjectAgent-based modelling
dc.subjectSurvival Analysis
dc.subjectDiscrete Multi-Species Prey-Predator Model
dc.subjectArtificial Life
dc.subjectArtificial Ecosystem
dc.subjectEnergy Flow
dc.titleSimulation of Collective Intelligence of a Multi-Species Artificial Ecosystem Based on Energy Flow
dc.faculty.departmentSciences des systèmes / Systems Science
dc.contributor.supervisorLee, Won-SookÉtudes supérieures / Graduate StudiesÉtudes supérieures / Graduate Studies
uottawa.departmentSciences des systèmes / Systems Science
CollectionThèses, 2011 - // Theses, 2011 -

Asgari_Aliakbar_2014_Thesis.movSystem Visualization - Simulation of Collective Intelligence of a Multi-Species Artificial Ecosystem Based on Energy Flow 3.5 MBVideo QuicktimeOpen
Asgari_Aliakbar_2014_Thesis.pdfSimulation of Collective Intelligence of a Multi-Species Artificial Ecosystem Based on Energy Flow 5.07 MBAdobe PDFOpen