Genetic Programming Algorithms for Dynamic Environments

Evolutionary algorithms are a family of stochastic search heuristics that include Genetic Algorithms (GA) and Genetic Programming (GP). Both GAs and GPs have been successful in many applications, mainly with static scenarios. However, many real world applications involve dynamic environments (DE). Many work has been made to adapt GAs to DEs, but only a few efforts in adapting GPs for this kind of environments. In this paper we present novel GP algorithms for dynamic environments and study their performance using three dynamic benchmark problems, from the areas of Symbolic Regression, Classification and Path Planning. Furthermore, we apply the best algorithm we found in the navigation of an Erratic Robot through a dynamic Santa Fe Ant Trail and compare its performance to the standard GP algorithm. The results, statistically validated, are very promising.