Analysis of large GPS trajectories datasets via multi-agent techniques

The widespread use of mobile devices and Internet services on users’ locations offer the opportunity to acquire information related to users’ trips in real time. This availability has given rise to several studies based on geospatial trajectories but, because of the large volume of the collected information, processing them together is usually difficult. In this thesis we illustrate an approach that, using a multi-agent system, provides personalized recommendations of Points Of Interest (POIs), such as libraries, museums, restaurants, etc. to users. In our context, an agent is an application that improves user navigation in a city. It collects opinions, in terms of scores, that quantify the level of satisfaction in visiting a certain place in a certain period of time. In this approach, interesting positions emerge from the analysis of the collected data, hence scores and suggestions may be available for any large city in any place, when a sufficient number of people provide data. In addition, the next places to visit are suggested according to people’s behavior and preferences. Other directions explored are the identification of the flows of multiple users and the intention to predict the paths that will be taken by a user on the basis of the common paths, already known, of other individuals. Given a large dataset of geographic trajectories in an urban metropolitan area, an efficient strategy for detecting corridors is also proposed. These can be defined as geographical paths, of a minimum length, commonly crossed by a minimum number of different users. This approach is important for transportation analytics because it allows to detect missing lines in public transport systems and also to advise private users which route to take to move from one part of the city to another based on the behavior of users who have provided their GPS logs. Although people like to visit popular places, due to health problems and due to the recent restrictions currently in place around the world for covid-19 influenza pandemic, meetings should be avoided. When planning a trip, one must consider both the attractiveness in terms of general interest for the destinations and the density of people who gather there. In the final part of the thesis, we propose a recommendation system that aims to offer to users some suggestions on useful routes and destinations which balance liveliness and overcrowding. First, we use datasets that store GPS locations as the basis for route and destination statistics. Then, we use an accurate probability algorithm that estimates the number of people moving from one place of the city to another and consequently show a list of destinations to users. Destination points are filtered according to the user’s preferences on the density of people. A multi-agent system is used to manage user’s requests to find a route for a journey, statistics on possible destinations and suggestions for users. Thanks to our solution, we can inform users about suitable routes and destinations, as well as alert them when a favorite destination is overcrowded.