Hierarchical Reinforcement Learning for Energy-Efficient Urban Planning

Urban planning plays a pivotal role in fostering sustainable, energy-efficient communities amidst the escalating challenges of climate change and energy crises. However, traditional methods often overlook the hierarchical interdependencies inherent in urban structures, land use configurations, and building design, leading to suboptimal outcomes. Here, we introduce a Hierarchical Energy-Efficient Urban Planning (HEEUP) framework, leveraging hierarchical reinforcement learning (HRL) to integrate decision-making across three tiers: urban structure, land use configuration, and building design. Our approach facilitates collaboration between cascading agents, enabling energy-conscious planning decisions that align with functional and contextual urban demands. Evaluations across real-world datasets from Miami and Albuquerque demonstrate a 7.3% improvement in energy efficiency compared to baseline methods. Furthermore, HEEUP achieves an average 24.52% reduction in computational training time compared to traditional optimization algorithms and generative machine learning models. These findings underscore the potential of HEEUP to significantly advance energy-efficient urban planning, providing a scalable, adaptable paradigm for addressing contemporary sustainability challenges.