In order to address the problem of large computation and high energy consumption in the process of multi-UAV collaborative missions, a multi-UAV adaptive task offloading scheme is proposed based on the computational offloading principle and game theory. The first step in the scheme is to model the system and construct a multi-node, mutually constrained mobile offloading model. Then, based on the unloading model, the method is constructed to calculate the time delay and energy consumption of the UAV when it performs its tasks respectively. The global cost function of the system is generated by considering both delay and energy consumption. An adaptive task offloading algorithm is designed based on the game theory and Nash equilibrium. The selection of the optimal computation node is achieved for minimizing the cost of the entire live system, thus balancing the UAV computation delay and energy consumption. Finally, compared with the existing models of offoading, task execution process of the proposed method has strong mobility, lower energy consumption and higher efficiency. The simulation results verify the effectiveness of the theory, which has a practical significance.