| dc.description.abstract | Nowadays large cities deal with serious problems in trac management. Those problems demand urgent expansion of the road network. However, usually it is not feasible due to the lack of space or due to high costs involved. Therefore, a trac engineering that provides fast and ecient solutions becomes essential. Among the possible strategies, the trac light programming optimization arises as an interesting choice, since it is an eective and low-cost solution. Several studies propose to optimize the trac light programming by maximizing the average speed of vehicles on road networks. However, many of these works use trac simulators to evaluate candidate solutions, which can become a problem to the optimization process. These simulators require high runtime, which prevents the optimization of trac light time in real time. Furthermore, the obtained solutions are evaluated for a single vehicle ow scenario, which causes doubts about its behaviour in the real world, in which vehicle ow can varies signicantly along the day and from one day to another. Because of these limitations, the aim of this study is to propose two new architectures, which use deterministic mathematical models, in order to obtain robust and optimized trac light programming in real time. Theseapproaches aim to minimize the time spent by vehicles to travel on network. Experiments were performed using mono and multi-objective optimization algorithms for real scenarios and they indicated the success of such architectures for this purpose. | |
