The article proposes an approach to adaptive traffic light management based on the reinforcement learning method and neural networks to improve traffic efficiency. The use of deep reinforcement learning allows the system to learn how to optimize traffic light phases, minimizing delays and queue length. The model was trained on data obtained from a traffic simulator, which increases its adaptability to dynamic traffic conditions. The efficiency of such a controller is evaluated by two different performance indicators to minimize the total queue length and total stop delays. The results show a significant improvement in reducing congestion and increasing throughput compared to traditional methods.