Document Type : Research Article
Faculty of Civil and Environmental Engineering, Tarbiat Modares University, Tehran, Iran.
Spatial analysis and distribution are of great importance to transportation planners, especially in traffic demand management. Simulation is an important tool in the planning and management of transportation systems to achieve an estimation of real system behavior to evaluate different scenarios. Regarding the aggregate nature and inability to consider heterogeneity among the individuals in a large number of discrete choice models and the high cost of data collection through questionnaires, using a disaggregate and heterogeneous agent approach can be used to evaluate different policies. Since each agent is inherently autonomous and interacts with different agents and the environment to achieve its goals, this paper aims to use the agent-based approach to simulate the destination choice of discretionary tours of Qazvin citizens. Individual socioeconomic characteristics and travel information questionnaires (revealed preference) of 9938 households and 29840 individuals in 12 municipality districts of Qazvin were collected. After extracting 12 types of activity patterns including shopping and recreation trips, the simulation of destination choice in MATLAB has been studied using the Reinforcement Learning algorithm (RL) and reward-punishment functions which are based on the relative attractiveness of districts for various modes and travel times. High correlation (above 0.9) results were achieved among simulated trip destination choice distributions and observed survey data using the RL algorithm which illustrates the algorithm's goodness of fit; also the simulation results and survey data have a similar trend among districts which illustrates that the simulation findings have real-world implications.