A swarm-based task allocation scheme is proposed to address the challenge of strike target allocation under degraded group communication conditions. This approach leverages guide head observations to construct an airborne model and define observation fields. By utilizing UAVs’ state information within local observation ranges as initial conditions, a reward function based on swarm density distribution is designed. The allocation process employs the MADDPG algorithm to effectively match targets with the swarm amidst communication degradation. Simulation results highlight rapid convergence and successful swarm allocation to multiple targets despite adverse communication conditions.