The governance of large tourist flows attracted by the historical cities is becoming a fundamental issue for the preservation of cultural heritage and the
life quality of residences. The Venice case study paradigmatic since the major has recently proposed limitations in the number of entrances in the historical center. Moreover the continuous increasing of the tourist flows corresponds a to a parallel decreasing of the number of residences changing Venice is a 'museum city' with large desert areas.
The main problems are:
1) to quantify the tourist flows understanding both the mobility demand related to the main attractive points of interest and the existence of critical congested situations;
2) to reconstruct the pedestrian dynamics on the whole road network pointing out the existence of preferred paths and statistical laws for the flow dynamics;
3) to build a dynamical stochastic model for the pedestrian mobility that allows short term predictions based on a real time data collection.
We propose to cope with the previous tasks by a complex systems physics approach based on ICTs data and a model inspired by the random walk on networks.
In a collaboration between the University of Bologna and the TIM S.p.a. mobile phone italian company we have analyzed the GPS data of a sample of $5\%$
of the whole population of the mobile phones present in Venice during two large tourist events: the Carnival and the 'Festa del Redentore' 2017. Moreover
another data collection campaign is planned during May 2018 using video-camera installed along the main paths connecting the railway station
and the Piazzale Rome parking to San Marco square for the people counting and video-cameras installed in San Marco square for the crowd counting (collaboration with CANON Italia) and performing a data fusion with the mobile phone distribution in the historical center of Venice. We will illustrate the main results of these 'experiments' from the point of view of both data analytics using deep learning neural networks and crowd counting algorithms and the possibility of modeling pedestrian dynamics on a road network to perform a forecasting of critical congested situations.