With the diffusion of interconnected sensors, systems, and stakeholders, smart cities are evolving into complex, data-intensive environments where urban data has become a critical asset. New challenges emerge in urban dataspaces regarding enforcing multi-actor, heterogeneous, dynamic and real-time policies. To explore the potential of applying Knowledge Graph (KG) technologies in the urban context, we identified a comprehensive list of real-world policy challenges related to data governance through a novel urban dataspace testbed and an illustrative use case. Among the key challenges are new categories of policy conflicts in urban data, complexity in multilateral data governance, and real-time sensor compliance scenarios. Key features are proposed to extend an existing Open Digital Rights Language (ODRL)-based policy engine to address these challenges, exploiting KG technologies and capabilities of large language models, including a concise formal semantics for ODRL, and enhancements to key components of the policy engine. We also provide concrete examples to illustrate the challenges and how this enhanced ODRL-based framework addresses them in the smart cities setting.