Genetic Algorithms in Search, Optimization and Machine Learning
Genetic Algorithms in Search, Optimization and Machine Learning
Co-evolving predator and prey robots
Adaptive Behavior - Animals, Animats, Software Agents, Robots, Adaptive Systems
| Hi-index | 0.00 |
One of the open questions in evolutionary computation is how an arms race may be initiated between coevolving entities such that the entities acquire new behaviors and increase in complexity over time. It would be highly desirable to establish the necessary and sufficient conditions which lead to such an arms race. We investigate what these conditions may be using a model of competitive coevolution. Coevolving species are modeled as points which are placed randomly on a two-dimensional fitness landscape. The position of the species has an impact on the fitness landscape surrounding the phenotype of the species. Each species deforms the fitness landscape locally. This deformation, however, is not immediate. It follows the species after some latency period. We model evolution as a simple hill climbing process. Selection causes the species to climb towards the nearest optimum. We investigate the impact different conditions have on the evolutionary dynamics of this process. We will see that some conditions lead to cyclic or stationary behavior while others lead to an arms race. We will also see spontaneous occurrence of speciation on the two-dimensional landscape.