Path Finding Algorithms for Autonomous Robots Based on Reinforcement Learning

449 www.ijarcet.org  Normally, the problem of path finding is solved by classical algorithms, such as the Dijkstra, Bellman-Ford,Johnsonalgorithm. However, the classical path finding algorithms cannot handle the problem of dynamic environment in the realworld.This paperintroducesmodelforautonomous robot path finding based onreinforcement learning algorithms.We describe the basic concept, principle and the method of reinforcement learning algorithms, and do some simulation experiments in order to evaluate the effectivenessamong different reinforcement learning algorithms.Simulation results show that Q-learning algorithm does not work well when the environment has changed constantly and there are too many obstacles.Both SARSA algorithms and Q-learning algorithms are relatively slow in convergence.Meanwhile, Dyna-Q algorithm determines the path of the robot relatively fast and the stability rates are high.