The Rise of Deep Generative, Hybrid, and Foundation Models for Sustainable Smart City Intelligence

This chapter lays the theoretical and practical groundwork for understanding how generative artificial intelligence (GenAI), particularly deep generative models, hybrid architectures, and foundation models (FMs), is reshaping the landscape of environmentally sustainable smart cities. It begins by contextualizing the book within the current technological transition toward generative intelligence as a new frontier of urban computing, emphasizing environmental sustainability as a core urban agenda. It then provides a conceptual framework for sustainable smart cities in light of recent advancements in GenAI and generative artificial intelligence of things (GAIoT). Moreover, it synthesizes recent review studies, categorizes key model families—generative adversarial networks (GANs), variational autoencoders (VAEs), diffusion models (DMs), and transformers, and hybrid models—and details their capabilities, trade-offs, and relevance to environmental planning, computational design, and climate resilience. It also addresses hybrid architectures that integrate generative, discriminative, deterministic, probabilistic, and/or heuristic approaches for complex urban and environmental systems and emerging urban and architectural design practices. Furthermore, it explores how deep generative and FMs are increasingly powering the next frontier of urban digital twins (UDTs), transforming them into intelligent and adaptive, and self-improving systems for sustainability-oriented, climate-responsive, and future-ready smart cities. Finally, it discusses how urban, environmental, climate, and geospatial FMs extend the power of GenAI in data-scarce, data-imbalanced, and high-uncertainty domains.