Study Of Intelligent Trajectory Planning Approach For Hypersonic Vehicles

Hypersonic aircraft is gradually becoming a research hotspot in the field of national defense and military industry due to its characteristics of long range,high speed,and strong maneuverability.This paper focuses on the reentry trajectory planning of hypersonic aircraft,based on the comparison of different intelligent planning methods and flight mission analysis,trajectory design based on supervised learning,trajectory planning based on model prediction static programming,and trajectory tracking based on adaptive parameter tuning based on reinforcement learning are carried out.The method of reentry trajectory planning based on supervised learning is studied.1)In response to the robustness requirements of neural networks,the correlation between the spectral norm method and the L2 regularization method is proved.2)When training samples are generated,the bank angle correction method in the numerical prediction correction algorithm is designed to reduce the unnecessary trajectory jump during the reentry process.3)A neural network learning single-step integration strategy is proposed,which realizes the rapid planning of the reentry trajectory and reduces the calculation time of the integration process.4)The trajectory planning task is transformed based on the pole-transformation model,which effectively reduces the complexity of input and output description in neural network training,and achieves higher training accuracy in fewer iterations.The reentry trajectory re-planning method of model prediction static programming is studied.1)A motion model with energy as the independent variable is established and simplified accordingly.2)The parameterized model of the control is designed,and the initial trajectory that meets the initial target point is generated by the method of numerical prediction and correction.3)A reentry trajectory re-planning method based on model-predictive static planning is proposed.Through discrete state and control variables,the attack angle and bank angle control that meet the performance index are analyzed analytically,and the state of the new target point is satisfied.4)Considering typical process constraints,penalty items are introduced to improve performance,and the application of model prediction static programming re-planning method under process constraints is realized.Considering the needs of adaptive parameter adjustment,the trajectory tracking method of parameter adaptive tuning based on reinforcement learning is studied.1)Using the LQR tracking guidance method,the trajectory tracking control model and linear quadratic performance indicators are established.2)The Markov decision mathematical description of the reentry trajectory tracking problem is established,so that the state,action,and reward function in the reinforcement learning process correspond to the state and control of the aircraft.3)A strategy to solve the feedback coefficient in real time according to the deviation of the state quantity is proposed,and the strategy optimization algorithm in the reinforcement learning is used to realize the adaptive adjustment of the control parameter in the trajectory tracking process.4)The Monte Carlo simulation verification considering different deviation conditions is carried out,and the adaptive tracking of the reentry trajectory under large deviation conditions is realized.The thesis focuses on the research on the intelligent trajectory planning method for artificial intelligence "enabling" in view of the complex multi-constraint conditions and nonlinear dynamic planning environment,which can effectively improve the on-line autonomous decision-making ability of hypersonic aircraft trajectory,and has reference significance for the development of hypersonic aircraft intelligence.