Abstract: To effectively address the complex challenges of multi-objective optimization in logistics for unmanned aerial vehicle (UAV) cargo transportation, this study introduces an innovative algorithmic framework: multi-objective deep deterministic policy gradient–nondominated sorting genetic algorithm (MODDPG–NSGA2). This two-layer architecture is specifically designed to address key issues such as path length minimization, time compliance rate maximization, energy consumption reduction, and adherence to multiple constraints. Furthermore, it seeks to overcome a critical limitation of the traditional nondominated sorting genetic algorithm II (NSGA2)—its inability to dynamically adapt to environmental changes through a closed-loop system that integrates perception, constraint handling, and real-time strategy adjustment. The upper layer of the proposed architecture leverages an enhanced version of the NSGA2 algorithm. By incorporating an advanced nondominated sorting strategy alongside an elite reservation mechanism, this layer constructs a robust multi-objective optimization model. The result is a global initial Pareto optimal solution set that comprehensively considers objectives such as minimizing path length, maximizing time achievement rates, and reducing energy consumption. This ensures that the solutions generated are not only efficient but also balanced across all relevant criteria. In the lower layer, the MODDPG algorithm played a crucial role in adapting to dynamic environments. It continuously senses and processes real-time information about environmental conditions, load-balancing requirements, and other state variables. Using deep neural networks, the algorithm approximates the state-action value function, enabling it to make informed decisions on path adjustments. This capability allows UAVs to dynamically modify their trajectories in response to changing circumstances, ensuring effective replanning within the local dynamic environment. For instance, if unexpected obstacles or weather conditions arise during delivery, the MODDPG algorithm can quickly recalculate the optimal route while maintaining efficiency and safety. The synergy between these two layers creates a deep, interactive, and collaborative mechanism that addresses common pitfalls in optimization problems, such as global imbalance and local myopia. Global imbalance refers to situations in which one objective (e.g., minimizing path length) is prioritized at the expense of others (e.g., energy consumption), leading to suboptimal overall performance. On the other hand, local myopia occurs when algorithms fail to consider long-term consequences and instead focus narrowly on immediate gains. By integrating both global and local perspectives, MODDPG-NSGA2 achieves a more holistic and adaptive approach to multi-objective optimization. Experimental results highlight the significant advantages of the MODDPG–NSGA2 algorithm over its predecessors. Compared with the traditional NSGA2 algorithm, MODDPG–NSGA2 demonstrates a 20.86% improvement in the task timeliness standard rate, which measures how consistently deliveries meet scheduled deadlines. In addition, it reduces the distribution path lengths by 21.90%, thereby lowering travel distances and associated costs. Energy consumption also decreased by 21.90%, contributing to greater sustainability and operational efficiency. These improvements are particularly pronounced under burst scenarios, in which sudden changes in demand or environmental conditions require rapid adaptation. To further validate the robustness of the proposed approach, this study conducted comparative analyses of five classical multi-objective optimization algorithms. The findings confirm that MODDPG–NSGA2 outperforms baseline methods by an average of over 30% in optimizing paths and reducing energy consumption. Moreover, it exhibits superior cross-modal optimization capabilities, meaning that it can effectively balance competing objectives, even in highly complex and unpredictable scenarios. This advancement has significant implications for enhancing the transportation efficiency of logistics for UAVs in urban environments. In densely populated cities, where traffic congestion and airspace restrictions pose significant challenges, the ability to dynamically and efficiently optimize routes can lead to substantial cost savings and improved service quality. Beyond logistics, the proposed framework offers a novel perspective for applying multi-objective optimization algorithms in dynamic complex systems. Its adaptability and scalability make it applicable to a wide range of domains, from autonomous vehicle navigation to resource allocation in smart grids. In summary, the MODDPG–NSGA2 algorithm represents a major step forward in addressing the multifaceted challenges of the logistics for UAV cargo transportation. By combining advanced optimization techniques with real-time adaptability, it provides a powerful tool for improving efficiency, reducing costs, and enhancing resilience in dynamic environments. As technology continues to evolve, this framework serves as a foundation for future innovations in multi-objective optimization and intelligent decision-making systems.