Multi-objective optimization for smart cities: a systematic review of algorithms, challenges, and future directions.

Area of Science:

• Urban Systems Engineering
• Computational Optimization
• Artificial Intelligence

Background:

• Smart cities increasingly rely on complex, interdependent systems for planning and decision-making.
• Multi-objective optimization (MOO) is crucial for enhancing smart city sustainability and real-time operations.

Purpose of the Study:

• To systematically review and classify MOO techniques applied in smart-city contexts from 2015-2025.
• To benchmark algorithm performance across diverse urban domains.
• To identify research gaps and propose a roadmap for future MOO frameworks.

Main Methods:

• Systematic literature review of 117 peer-reviewed studies.
• Classification of MOO algorithms into four families: bio-inspired, mathematical, physics-inspired, and ML-enhanced.
• Benchmarking based on efficiency, scalability, and suitability for six urban domains (infrastructure, energy, transport, IoT, agriculture, water).

Main Results:

• Established MOO algorithms like NSGA-II and MOED/D are prevalent.
• Hybrid frameworks combining deep learning and evolutionary search show enhanced adaptability in dynamic, high-dimensional smart-city environments.
• Key challenges identified include limited cross-domain generalizability, poor uncertainty handling, and low interpretability of AI-assisted models.

Conclusions:

• Hybrid MOO approaches offer significant potential for smart-city applications.
• Addressing research gaps in privacy, trade-off resolution, digital twin integration, LLMs, and neuromorphic computing is essential.
• A benchmarking toolkit and algorithm-selection matrix are provided to guide practical implementation and future research in scalable, interpretable, and resilient urban optimization.