In order to achieve continuous remote communication between multiple source nodes and the Ground Control Station (GCS), a solution method is proposed for dynamic path planning of relay unmanned aerial vehicles based on the artificial potential field method. The relay cluster computation module is to dynamically divide the relay clusters according to the maximum density clustering, realizing the dynamic adjustment of the number of relay nodes. The dynamic path planning module for relay nodes employs in accordance with the concept of artificial potential fields, enabling the relay drones to mobilize to suitable positions based on the resultant force of the potential field while possessing dynamic obstacle avoidance and collision prevention capabilities. The simulation experiments are conducted on the constructed model with different task payload settings. The results demonstrate that the relay unmanned aerial vehicles can reliably achieve multi-hop relaying between multiple source nodes and the GCS, with a connectivity rate of communication links being no less than 99.83%, the number of source nodes being no greater than 10, and the number of threat zones in the airspace being no greater than 7.