Optimal metro route identification using Q-Learning

Public transport plays quite an essential role in the life of a common man. In metropolitan cities, people try to avoid road-based public transport due to reasons like increasing road congestion, air pollution, energy, and oil consumption, and most importantly to save time. Metro is the most common means of transportation which connects the whole city and the nearby areas via a train that runs on various routes connecting every station in the city. The aim of this paper is to find an optimal metro route for a passenger. The proposed system finds all the stations in order the passenger needs to visit or pass by to reach his destination station by the shortest route. The system considers the whole Metro Map as a graph where the stations act as nodes and are connected using edges. The system uses the Q-learning algorithm to find the optimal path. Q-learning is a reinforcement learning algorithm that teaches an entity which actions to perform under which conditions. It is a reward-based algorithm, stateless and off-policy and hence learns how much long-term reward it will get for every possible action taken on or from every state. Q-learning is also a feedback-based algorithm, it carries forward the output of the previous step as input to the next step. The proposed technique depicts the optimal metro route found by the algorithm. The proposed system aims to reduce the idle/stationary time by reducing the number of stations traversed while reaching the destination station which in turn reduces the total travel time.