Modeling the behavior of mobile robots using genetic algorithms

Abstract

The article is devoted to the analysis of the behavior of a mobile robot using finite state machine algorithms in order to find a way to the goal and avoid obstacles. After justifying the use of such methods, the analysis of a standard deterministic finite automaton is done. Further, the theory of Markov processes is applied to this algorithm, as a result of which the state machine becomes part of the hidden Markov model. This allows you to apply probabilistic methods to modeling the behavior of the robot. This probabilistic behavior is most promising in complex environments with unpredictable obstacle configurations. To compare the efficiency of deterministic and probabilistic finite state machine, we applied a genetic algorithm. In the numerical experiment that we conducted in the Scilab software, we considered two main types of environments in which a mobile robot can move - an office-type environment and a polygonal-type environment. For each type of environment, we alternately applied each of the indicated behavior algorithms. For the genetic algorithm, we used one hundred individuals who were trained over 1000 generations to find the most optimal path in the specified environments. As a result, it was found that the deterministic finite state machine algorithm is the most promising for movement in an office-type environment, and the probabilistic finite state machine algorithm gives the best result in a complex polygonal environment.