The Optimal Path Finding Algorithm Based on Reinforcement Learning

Introduction

Smart city concept has been evolved in recent years. 2007 was the first year where the trend has been changed. People situated in urban areas increased as compared to rural. As per the United Nations report in 2018, approximately fifty-five percent of people in the world live in urban areas (Glaeser, 2020). By the end of 2050 sixty-eight percent, people are expected to be urbanized. Urbanization rapidly increasing in Asia and Africa region as compared to others, with India and China making the main difference. The trend of urbanization affected all the things right from transportation, smart cities, health facilities, automation, etc. The number of vehicles around the world is also increasing briskly. Currently, 1.2 billion vehicles are running on the roads and by the year 2035, it would become 2 billion. An increase in vehicles (Guidoni et al., 2020) creates problems like congestion, traffic jams, accidents, parking issues, pollution, economical investment in infrastructure building, etc.

An intelligent traffic system (Zhang et al., 2019) comes with intelligent algorithms for optimal pathfinding. In case of emergency situations drivers can’t take decisions accurately (Jhiduk, 2014) but with the help of intelligent algorithms they can avoid situations like accidents, jams, roadblock, etc. that’s why the novel algorithm proposed in this article can intelligently provide the optimal path considering multiple parameters. The route planning of the intelligent path algorithm is based on the current traffic scenario by considering history data, cost, etc. It is based on the continuous feedback taken from various resources and the optimal path intelligently determining method. It is partitioned into two ways, local route planning, and global route planning (Zhang et al., 2018). Global route planning works on the concept of optimization and survey providing method using local data in case of a well-known data house to know the possible areas and optimal routes (Zhang et al., 2018). As route provided by the global route generator is not the final one as it does not consider multiple objects during calculations. So the local route generator must be based on local traffic data (Zhang et al., 2019). The local and global route generator algorithms are correlated. In some cases, local information is gathered with the help of sensors and scenario is analyzed (Zhang et al., 2019). The static algorithms like A* algorithm are the quickest ones to provide an intelligent optimal route in state space. But static algorithm will not serve the purpose as traffic scenario is dynamically changing. So we need to switch on other technological algorithms. Reinforcement Learning is a vital innovation of upcoming Machine Learning technology. Reinforcement learning provides the facility with the machine to take the decisions the way humans do (Ren, 2017). The main advantage of using reinforcement learning is to provide long term solutions by considering the current state scenarios. It takes into consideration the environmental updating and can learn from the errors made in the past (Zhang et al., 2019). The optimal route for a traveler can be provided by considering multiple parameters like minimum distance, weight, total journey time, various performance parameters, etc. In this article, the novel dynamic algorithm for optimal pathfinding has been proposed which is based on reinforcement learning with multi objectives parameters. By the end of the year, 2050 the number of vehicles will become double i.e. 2.5 billion. In the metro cities like Bangalore in India, people spend more than half travel time in jams waiting for traffic roads to be clear. Sometimes even google map doesn’t show any road is blocked due to repairing work, accidents, jams, etc. The proposed algorithm provides an optimal path to the driver within a fraction of second so that traveling can be done with ease, comfort, in minimum time and with the shortest distance. It also overcomes the problem that arises due to sudden dynamic changes in traffic. It will reduce traffic jams, accidents, fuel consumption perhaps pollution.