Communication and machine learning in urban mobility: a multiagent and multiobjective approach

Abstract

Machine learning is being used in the area of urban mobility. In particular, reinforcement learning (RL) is a technique that is frequently used since it allows agents to adapt to the state of the traffic. However, the literature on RL rarely discusses works that tackle multiobjective decision making, that are key in this domain. This way, the goal of this project is to develop a framework, as well as methods to take into account multiobjective decision making when agents of various types (drivers, traffic signals) are deciding what actions to take. In order to improve the efficiency of the learning process, we propose the use of transfer learning techniques. To this aim, new methods will be developed, using car 2 car and other forms of communication. (AU)