Physics of the city and urban mobility models
- Introduction - B. Giorgini Modelli
- Models, measurements and tests - G. Melchiorre
- Distrimobs: a design tool - M. Brambilla, L. Cattelani
- Individual behaviours: the decision-making mechanism - O. Bernardi
- Results and prospects - S. Rambaldi
Introduction - B. Giorgini
It presents urban mobility as a complex phenomenon that can be explored by modelling virtual systems composed of "intelligent" individuals that explore in silicon the potential features and dynamic changes with emerging properties, such as self-organisation. In other words, an artificial intelligence for mobility is modelled. The tricky point is how the artificial world with its agent population can reflect the reality of its population of citizens. Certainly, the problem of collecting time-extended series of empirical data to be compared with the simulation results is crucial for the future development of this type of models.
This introduction presents the work done by AIMLab, Artificial Intelligence for Mobility, a laboratory which is set on a research program to study mobility all over Bologna and which, with Laboratorio Nomadis of Milan's Università Milano Bicocca, is analysing the empirical data of pedestrian mobility and congestion phenomena during the Venice Carnival.
Finally, the latest frontier in the study of traffic is briefly mentioned.
Models, measurements and tests - G. Melchiorre
It addresses the issue of personal mobility, including both pedestrians and motor vehicles, which can be described by means of physical-mathematical models, and, for them not to remain purely theoretical, they must be explored through some real cases after appropriate adjustments, while they must be tested in order to highlight any discrepancy or inconsistency between the models and reality.
It presents several models, their adjustments and testing, while highlighting the specific features of each one: each model has been designed to reproduce a given type of mobility on a given space scale.
The pedestrian mobility that can take place within a confined space, for instance a square, a museum, a station, is illustrated by the Campus model. The mobility of motor vehicles, which can take place within a complex grid, such as a metropolitan area, is illustrated by AutoMobilis. Finally, the Mileto/Manhattan model illustrates a mobility which combines pedestrian mobility with public transport and private motor vehicle mobility.
Calibrating such models means to determine the different populations that are typical of each type of mobility on a real case and define the main space-time relations of the tested area, which can range from a station kiosk to a shopping mall in a large metropolitan area.
Testing means measuring the paths that define the main tracks of the area and comparing them with those reproduced by the models.
Distrimobs: a design tool - M. Brambilla, L. Cattelani
Simulators are helpful tools for the computer-aided design of traffic and mobility. This presentation introduces the urban design tools, starting from those that have been historically used by public bodies and showing the main features of each type of tools. This is followed by a review of the Distrimobs software project which is being developed at AIMLab, based on an agent-type architecture.
Then, it introduces the agents of the Distrimobs model, the mobbers, which stand out for individual pedestrians and, just like them, are provided with perceptions (detailed perceptions from the short distance and aggregated perceptions from a long distance) and a memory. In particular, it emphasises how each mobber acts to fulfil its own motivation, which change over time, based on the currently-perceived state of the environment. Finally, it shows the workings of the mobbers' layered artificial intelligence, in which the lower layers take local decisions, while the higher layers implement far-ranging policies.
The Distrimobs software is then put forward as an urban and architectural design tool by showing how it offers the possibility of making adjustments to existing projects from the most common formats (CAD, ESRI Shape...). This paper shows that, importantly, Distrimobs can simulate a high number of agents using its own distributed architecture. Finally, through a visual exploration tool, it shows how simulation results can be analysed and presented.
The speech is ended by some examples of applications of the simulator, with special reference to the innovative features of the Distrimobs software and a few real applications for the near future...)
Individual behaviours: the decision-making mechanism - O. Bernardi
This presentation speaks of the fact that, since the elementary components of the city as a system, the individuals, exchange information with the environment all the time and have cognitive skills, in order to describe this system, an "intentional and decision-making physics" is needed.
Specifically, it presents a mechanism for making a choice between two possible destinations after receiving information about their congestion. What can be observed is the appearance of a cooperative phenomenon, which leads to a collective dynamics, whereby the individuals are inclined to a self-consistent balance, thus bringing the system to a stable, self-organised configuration.
This paper shows that this phenomenon is determined by the parameters introduced in the model, such as the time it takes to collect the information, the capacity of the visited places, the level of cooperation of each individual, which determine a different trend in the decision-making process of the individual (the microscopic level) and the self-organised or non self-organised configuration of the population (the macroscopic level).