This paper presents a hybrid deep learning framework designed for real-time adaptive navigation in dynamic urban environments. Our model combines VGG16, a robust convolutional neural network, with a Vision Transformer (ViT) to enhance the decision-making capabilities of autonomous vehicles by efficiently processing visual data. Training was performed on a diverse dataset of 30,900 images generated by the VSim-AV simulator, which emulates various urban scenarios. Experimental results indicate that the proposed framework achieves an accuracy of 98% and a precision of 96.3%, with a rapid processing time of 0.017 s per decision. These findings demonstrate the model’s potential to deliver reliable and timely navigation in complex real-world settings. By merging the strengths of traditional CNNs with modern transformer architectures, this study contributes a novel approach to autonomous vehicle navigation, underscoring the importance of simulation in advancing adaptive navigation systems.