The rapid advancement of UAV technology has elevated the strategic significance of multi-UAV cooperative systems in modern warfare. However, multi-source uncertainties in complex battlefield environments significantly increase the complexity of task allocation. To address this, this paper proposes study on UAVs cooperative task allocation problem based on the improved Genetic Algorithm, aiming to address the challenge of efficient decision-making under high-dimensional uncertainties. The paper begins by constructing a joint uncertainty set that integrates space, time, and resource dimensions. It then employs robust optimization to model uncertainties in a unified manner. This approach effectively mitigates the curse of dimensionality associated with high-dimensional uncertainties. The paper presents three innovations at the algorithmic level. First, it designs an adaptive mutation rate strategy based on the Sigmoid function to dynamically balance global exploration and local exploitation. Second, it introduces the dynamic crossover operator to adjust the crossover probability and method in real time based on population diversity, thus improving the convergence efficiency. Third, by combining Monte Carlo robust evaluation with importance sampling and segmented weight, it quantifies anti-interference ability and optimizes both cost and robustness in stages. Finally, the paper validates the model’s performance through multi-dimensional simulation experiments. The results demonstrate significant advantages of the proposed model over the traditional Genetic Algorithm and the PSO Algorithm. Specifically, the model shows improvements in solution quality, anti-interference ability, and dynamic response efficiency. Consequently, the study successfully constructs UAVs cooperative task allocation problem based on an improved Genetic Algorithm that integrates efficiency-robustness-dynamics.