The formation of UAVs is studied by introducing game theory into distributed control,and a game-based distributed algorithm is developed.The algorithm combines the gradient descent and consensus to convert formation of UAVs into a problem of distributed game.The consensus technique is adopted to guarantee the estimation of the positions of all UAVs,and gradient descent is for the Nash equipment.By the proposed algorithm,UAVs can complete any formation even each UAV only have the neighbor UAV's information not all UAVs'.The simulation result verifies the effectiveness of the proposed algorithm.