Calibrating a New Reinforcement Learning Mechanism for Modeling Dynamic Activity-Travel Behavior and Key Events

摘要: Recent transportation modeling mainly focuses on activity-based modeling. However the majority of such models are still quite static. Therefore, current research aims at incorporating dynamic components, as short-term adaptation and long-term learning, into these models. In particular, this paper attempts simulating learning process underlying development activity-travel patterns. Furthermore, study explores impact key events generation daily schedules. The algorithm implemented in uses a reinforcement technique, for which foundations were provided previous research. goal present is to release predefined sequence assumption allow determine autonomously. To end, decision concerning transport mode needs be revised well, aspect was previously also set within fixed schedule. order generate feasible patterns, another alteration consists time constraints, example opening hours shops. addition, event, case “obtaining driving license”, introduced methodology by changing available modes. resulting patterns reveal more variation selected activities respect imposed constraints. Moreover, observed dissimilarities between schedules before after event prove significant based alignment distance measure.