Real-World Evaluation of Hybrid Green AI for Sustainable and Efficient Smart Supply Chain Distribution

Optimizing physical distribution in smart supply chains requires a balance between efficiency and environmental sustainability. This paper proposes a hybrid green AI framework, integrated with real-world geospatial data, to achieve multi-objective logistics optimization. The framework combines metaheuristics and multi-objective optimization (MOO) algorithms to identify delivery routes that minimize costs, time, fuel consumption, and CO2 emissions while maintaining operational efficiency. The system uses Google Maps directions to obtain accurate distance and travel time matrices, ensuring that routing decisions reflect actual road networks, traffic, and terrain constraints. The optimization layer uses OR tools to prototype multi-objective optimized ant colony (MOIAC) and multi-objective optimized particle swarm (MOIPS) algorithms, enabling scalable and efficient computing. Dynamic demand patterns are simulated using Zipf stochastic distribution models, allowing for adaptation to irregular customer demand across urban and rural nodes. Experiments conducted in 19 Egyptian cities showed that the MOIAC algorithm reduces total distance, operating costs, and carbon emissions by 26.7% compared to baseline methods. It also achieves the lowest average response time (120 ms) under highly skewed demand (Zipf coefficient α = 0.9), outperforming other algorithms. The proposed framework integrates green AI, geospatial intelligence, and multi-objective optimization to provide a practical and sustainable solution for smart supply chain distribution, while balancing operational performance and environmental responsibility.