Addressing the issues of low transformation efficiency and poor stability during the dynamic transformation process of UAV swarms, this paper proposes a task-driven homogeneous swarm multi-geometric formation dynamic transformation flight control algorithm, by introducing a PD controller to improve the consensus control law and using the Hungarian algorithm to assign target positions to each UAV during the formation transformation. Finally, simulation and actual flight experiment were conducted for verification. The experimental results show that compared with the traditional consensus method based on Laplace model, the proposed method can converge the position error of the UAV swarm to the set threshold faster, improve the formation transformation speed by 80%, and achieve dynamic transformation of multi-geometric formations, meeting the requirements driven by complex tasks.