Q value-based dynamic programming with evolving penalties for road networks

Nowadays, many people depend on the guidance provided by car navigation devices to travel to the destination. Generally, several routes are possible for a given origin-destination (OD) in a road network. Often car navigation devices provide the minimum traveling time or distance route. However, this route may not always be comfortable for users to follow. Out of the several available routes, users may prefer to travel on a certain route depending on their preferences. In this paper, we propose a Q value-based Dynamic Programming with Evolving Penalties to find the user preferred optimal route. In the proposed method, we consider multiple criteria and take users' preferences for them. The users' preferences are used to determine the penalty values for the criteria, which are represented as the individuals of Genetic Algorithm. In the training phase, GA population is used for the evolution of penalty values of the criteria. The values obtained from the best individual is stored and used in the route search. The algorithm was applied to the road network of Fukuoka prefecture and results show that we can find the route considering users' preferences.