Deployment of Artificial Neural Networks for Vehicular Road Traffic Flow Prediction and Adaptive Signal Control in Metropolitan Road Networks

This study integrates ANN prediction with adaptive traffic signal optimization within a SUMO simulation framework for Nigerian urban traffic road networks, addressing the persistent challenge of urban congestion. Using synthetic and real traffic data from Jattu Junction in Auchi, Edo State, the framework was implemented with SUMO, Python, Pygame, and MATLAB for simulation, training, and evaluation. The ANN architecture featured an input layer, three hidden layers of 50 neurons each with Rectified Linear Unit (ReLU) activation, and a linear output layer, trained with the Adam optimizer (learning rate is 0.001, batch size is 64, with 100 epochs). Results showed that the adaptive system reduced average waiting time from 2.682 minutes to 1.579 minutes which is a 41.12% improvement. The ANN achieved strong performance metrics (MSE is 0.0154, MAE is 0.0923, R² is 0.9824), demonstrating its effectiveness in reducing congestion and improving mobility. Traffic scenarios included peak-hour demand, incidents, and varying flow conditions, with datasets split into training (70%), validation (20%), and testing (10%). Overall, the findings highlight the potential of ANN-driven intelligent transportation systems for enhancing urban traffic management.