The dynamic behavior of traffic signals controlling an isolated intersections can be seen as a discrete event system that can be modeled by Petri nets. The approach used in this paper proposes an adaptive control strategy of urban signalized intersection in real time using T-timed Petri Nets (T-TPNs). This strategy allows to manage the green light time autonomously according to the queue size for each flow in each lane of the intersection. A decentralized approach is proposed for the decision-making, where every phase in the intersections can be considered as a subsystem to be modelled and controlled (by traffic lights) separately, then a global controller allows to manage the conflicts between several phases.