Travel management optimization based on air pollution condition using Markov decision process and genetic algorithm (case study: Shiraz city)

Currently, air pollution and energy consumption are the main issues in the transportation area in large urban cities. In these cities, most people choose their transportation mode according to corresponding utility including traveller's and trip’s characteristics. Also, there is no effective solution in terms of population growth, urban space, and transportation demands, so it is essential to optimize systematically travel demands in the real network of roads in urban areas, especially in congested areas. Travel Demand Management (TDM) is one of the well-known ways to solve these problems. TDM defined as a strategy that aims to maximize the efficiency of the urban transport system by granting certain privileges for public transportation modes, Enforcement on the private car traffic prohibition in specific places or times, increase in the cost of using certain facilities like parking in congested areas. Network pricing is one of the most effective methods of managing transportation demands for reducing traffic and controlling air pollution especially in the crowded parts of downtown. A little paper may exist that optimize urban transportations in busy parts of cities with combined Markov decision making processes with reward and evolutionary-based algorithms and simultaneously considering customers’ and trip’s characteristics. Therefore, we present a new network traffic management for urban cities that optimizes a multi-objective function that related to the expected value of the Markov decision system’s reward using the Genetic Algorithm. The planned Shiraz city is taken as a benchmark for evaluating the performance of the proposed approach. At first, an analysis is also performed on the impact of the toll levels on the variation of the user and operator cost components, respectively. After choosing suitable values for the network parameters, simulation of the Markov decision process and GA is dynamically performed, then the optimal decision for the Markov decision process in terms of total reward is obtained. The results illustrate that the proposed cordon pricing has significant improvement in performance for all seasons including spring, autumn, and winter.