Agent Simulation of Time Allocation on Daily Activity-Travel Patterns

摘要

Time allocation is a crucial part in daily activity-travel patterns of individuals. An agent-based model is presented in this paper to simulate time allocation of typical activity-travel patterns based on travel diary survey data of Jiangning District, Nanjing City. Q-learning method which is one of the most useful ways to solve reinforcement learning problems is adopted in this paper. Several representative activity-travel patterns are first extracted from the data. Reward functions are then established according to the real-word data which increases our method more practicable. Transportation environment is also defined as a special agent and its influence on individual decision is measured by applying the U.S. Bureau of Public Roads (BPR) model. The simulation results indicate that in order to obtain total maximum reward, different activity-travel patterns have specific features on time allocation. Besides, simulation results are compared with the survey data to verify the accuracy of our model.