Gravitationally Inspired Search Algorithm for Solving Agent Tasks
Margarita Spichakova · Baltic Journal of Modern Computing · 2017
Artificial ant problem is defined as constructing the agent that models the behavior of the ant on trail with food.The goal of the ant is to eat all food on trail with limited number of steps.Traditionally, the recurrent neural network or state machines are used for modeling ant and heuristic optimization methods for example Genetic Programming as search algorithm.In this article we use Mealy machines as ant model in combination with Particle Swarm Optimization method and heuristic algorithms inspired by gravity.We propose new gravitationally inspired search algorithm and its application to artificial ant problem.Proposed search algorithm requires discrete search space, so the specific string representation of Mealy machine is introduced.The simulation results and analysis of the search space complexity show that the proposed method can reduce the size of the search space and effectively solve the problem.