Multi-Objective Monte-Carlo Tree Search based Aerial Maneuvering Control

This paper studies the problem of aerial maneuvering control, which is formulated as a multi-objective sequential decision making problem, and is solved by the Multi-Objective Monte-Carlo Tree Search (MOMCTS) algorithm which requires little prior knowledge about the specific maneuvering procedure, and achieves the desired maneuvering trajectory through multi-objective optimization. The proposed method is validated by simulation based on the six degree of freedom aerodynamic model of a real trainer aircraft in the case of loop maneuvering. The performance of MOMCTS approach is compared with a traditional method combining trajectory generation and PID control. Experimental results show that MOMCTS outperforms the baseline method at the cost of a lightly decreased robustness.