ARCHIVES

The rapid expansion of urban mobility has intensified the need for intelligent and dynamic traffic management systems which are proficient to respond for real- time situations. This study presents Intelli Flow, a novel framework for predictive traffic pattern optimization in smart cities, driven by the proposed Deep Q- Network with Genetic Algorithm Flow Optimizer (DQN-GA Flow). The hybrid model integrates deep reinforcement learning and evolutionary computation to achieve optimized signal control and efficient traffic flow management. The DQN component enables dynamic state-action mapping for continuous learning of traffic conditions, while the genetic optimization mechanism refines policy parameters to enhance convergence stability and global search performance. Simulation experiments were conducted using a benchmark traffic dataset within a controlled smart city environment, and comparative analysis was performed against three existing algorithms: Fuzzy Logic-Based Traffic Controller (FLTC), Q- Learning-Based Traffic Signal Control (QL-TSC), and Deep Deterministic Policy Gradient (DDPG). The evaluation metrics included Average Vehicle Delay, Network Throughput, Convergence Rate, and Average Travel Time. The results demonstrated that DQN-GA Flow achieved significant improvement in performance, recording reductions of up to 24% in travel time and 22% in vehicle delay, along with notable enhancements in throughput and learning efficiency. The findings confirm that hybrid reinforcement learning combined with genetic optimization provides an effective approach for achieving adaptive, data-driven, and sustainable traffic management in smart city infrastructures.