Today, more than 55% of the world’s population resides in cities, a figure that is expected to rise to 68% by 2050. This rapid growth in urbanization has made traffic congestion a global problem. This phenomenon severely affects urban mobility in many cities, hindering the daily commute of millions of people and generating negative economic, environmental and social impacts. This study explores the application of Deep Reinforcement Learning (DRL) as an approach for adaptive traffic signal control. The main objective is to develop a DRL-based system for traffic signal control capable of improving the traffic flow efficiency under varying traffic conditions. The proposed methodology includes the following phases: control problem definition, reinforcement learning environment, DRL model, and training. The model was evaluated through simulations on the Simulation of Urban MObility (SUMO) platform. The experimental results show that the DRL agent reduces the total cumulative waiting time by an average of 97.54% in low traffic conditions and 98.43% in high traffic scenarios. In addition, the number of vehicles within the system improved by an average of 74.86% under high traffic conditions. The study concludes that DRL offers a promising and scalable solution for real-world urban traffic management, even in very complex traffic contexts such as in Peru (Lima and Callao), which have some of the most chaotic traffic in the world. Future work should address multi-intersection coordination and the inclusion of diverse road users.