Applying Reinforcement Learning for Shortest Path Problem

Abstract

With the rapid development of high technology and artificial intelligence, there are plenty of novel software inventions have been created. Thus it is essential to consider how to make these innovations more practically to improve the convenience and efficiency of daily life. Therefore, this paper is concerned about implementing the machine learning method to address problems in daily life. Thus, a novel form of the reinforcement learning algorithm is applied to the shortest path problem abstracted from real life. The problem focuses on finding the most optimal route on a ten-note weighted graph from one point to a destination. The agent is trained four hundred times to accumulate values in the Q table, then the result is determined by the final Q value. Because of the success of resolving this problem, this algorithm should be implemented practically despite some limitations. With the help of this algorithm, the agent can determine the desired route extremely quickly to improve efficiency.