With the increasing application of heterogeneous unmanned platforms in complex urban environments, efficient path planning in dynamic scenarios has become a research hotspot in the field of intelligent unmanned systems. This paper proposes a dynamic path planning algorithm for heterogeneous unmanned platforms based on deep reinforcement learning, aiming to address challenges such as building obstructions, signal attenuation, and platform collaboration in urban environments. By constructing a multi-objective cost function encompassing building obstacles, communication strength, and task requirements, the algorithm enables optimized path selection in real-time dynamically changing environments. Furthermore, the proposed method integrates sensor data from drones and unmanned vehicles to enhance target detection during the path planning process, ensuring effective collaboration among platforms. Simulations and comparisons with existing path planning algorithms validate the superiority of the proposed approach in terms of path planning efficiency and task completion time, particularly demonstrating strong adaptability and real-time performance in complex urban scenarios.