Flight Test Mission Order Scheduling Approach Based on DQN
Luofu Wu, Lunhao Ju, Xinyang Wu, Guanren Chen, Hanqiao Tao, Guopeng Song · 2025
Since the twentieth century, the flight test field and its planning problems have garnered significant attention due to their inherent complexity and substantial potential benefits. In practice, the optimization method is rough, and the optimization effect is general. This paper focuses on the test tasks, innovatively formulating a linear integer programming model for the flight test mission order scheduling problem (FTMOSP). On this foundation, we further devise a solution algorithm based on the Deep Q-Network (DQN), aiming to tackling this intricate problem efficiently and intelligently. To validate the effectiveness of the proposed model and algorithm, we have expanded upon the benchmark for the bin packing problem with precedence constraints by introducing critical elements such as configuration constraints and trial point weights, constructing a dedicated case set applicable to the research problem at hand. Subsequently, using this case set, we conduct a comprehensive comparative analysis between the proposed DQN algorithm and a practical solution method, the Random First-Fit algorithm. Experimental results demonstrate that the proposed DQN algorithm exhibits superior efficiency and stability in solving the FTMOSP. This discovery not only offers new theoretical support and practical guidance for flight test planning but also provides fresh insights and methodologies for solving linear integer programming problems. The work presented in this paper holds scientific and practical value in advancing the in-depth study and application of flight test planning problems.