Interacting Agents with Memory in Virtual Ecosystems

Bedrich Benes, Javier Abdul Cordoba, Juan Miguel Soto
Department of Computer Science
ZIP code 14380; Mexico D.F.



Keywords: Virtual plant ecosystem, visual simulation, agents with memory, procedural modeling


An agent-based modeling of virtual ecosystems is presented. A virtual ecosystem develops by plant competition according to biologically inspired rules and tends to reach stability. Virtual agents enter the ecosystem and perform actions that favor certain plant species and cause system instabilities. Agents liberate space for some plant species by eliminating the others, they take out old plants if an area is overcrowded, they seed plants, water them, etc. Agents synchronize by message passing to cover the area efficiently and not to interpenetrate their areas of influence. Each agent has a local memory of pending tasks. When a memory overflow arises the agent divide the pending tasks among the nearest agents.