Urban mobility and logistics systems are increasingly challenged by congestion, rising fuel costs, and sustainability requirements. Traditional route optimization approaches, such as Dijkstra’s and A* algorithms, are limited in their ability to adapt to dynamic traffic conditions and to balance multiple conflicting objectives. This paper proposes an AI-powered multi-objective dynamic route optimization framework that integrates reinforcement learning, evolutionary algorithms, and graph-based learning to generate adaptive and sustainable routing solutions. The framework simultaneously minimizes travel time, fuel consumption, and CO₂ emissions, while maximizing on-time delivery performance. Real-time traffic data and logistics constraints are incorporated to enable continuous re-optimization. Experimental evaluation was conducted using both simulated city networks and real-world datasets for ride-sharing and logistics scenarios. Results demonstrate that the proposed framework outperforms baseline shortest-path algorithms, achieving up to 18% reduction in travel time and emissions, along with a 20% improvement in on-time delivery rates. The study highlights the potential of combining AI-driven adaptability with multi-objective optimization to enhance efficiency, reliability, and environmental sustainability in smart urban mobility. Limitations related to computational overhead, scalability, and data dependency are discussed, and future work directions include integration with edge computing, multi-modal transport systems, and city-scale pilot deployments.
Kumaresh et al. (Sat,) studied this question.