LA-Based Approaches to Infer Urban Structure from Traffic Dynamics Considering Costs

Successful future urban planning is highly dependent on optimal connectivity between important areas of cities. Discovering essential latent links will optimize the urban structure. Moving towards a better structure requires some information. There are a lot of sources of information for urban structure inferring, including the current structure, the time-varying traffic dynamics, and the construction costs, which are the basics of the optimization problem formulation. This paper presents a new formulation for the problem. The model problem to be solved tries to utilize all data sources needed for inferring. There are some methods for solving the formulated problem. The methods need some development to apply to the model. Methods utilizing learning automata (LA) are very favorable in this field due to the interaction with the environment. This paper presents two LA-based approaches for the model: Distributed Learning Automata (DLA) and Cellular Learning Automata (CLA). The algorithms result in an optimal connectivity matrix considering urban structure, traffic dynamics, and costs, where the matrix must include the current urban structure and some new reasonable necessary links. Moreover, comparisons are possible because the model has a fitness value for evaluating the provided connectivity matrix. The CLA-based proposed method performed better than the others in most experiments.