Aircraft Flight Autonomous Decision-Making Method Based on Target Predicted Trajectory and Markov Decision Process

In this paper, in order to enhance the autonomous operation capabilities of aircraft and ensure their operational safety and efficiency, we propose an autonomous decision-making framework based on target motion state prediction combined with the Markov Decision Process, namely IMM-MDP architecture. Firstly, our own aircraft utilizes the IMM algorithm to achieve the state prediction of the target; building upon the existing algorithmic structure, the proposed framework improves model sets within the IMM algorithm by incorporating climb and turn models and refines the motion modes of aircraft during the cruise phase, gathering corresponding sub-models to predict different motion modes and enhance the prediction accuracy of the target. Secondly, we adopt the Markov Decision Process as the autonomous decision-making method for the own aircraft, proposing a method by which to calculate the optimal decision sequence based on the prediction scenarios of the IMM algorithm; that is, in both multi-step prediction and single-step prediction scenarios, the payoff values of different action strategies at each decision moment are calculated to obtain the optimal decision sequence. The experimental results show that the IMM algorithm with the improved model set, described in this paper, is more accurate than the IMM algorithm and prediction results of the current statistical model, as described in the literature. We also set up a multi-aircraft operation scenario, comparing the IMM-MDP decision framework proposed in this paper with the Monte Carlo and MPC decision models, thus demonstrating that the proposed framework provides better decisions while ensuring safety. Due to the target state prediction and updating, this method also demonstrates better real-time performance and practicality.