Trajectory Design And Optimization Of Low-thrust Spacecraft Via Shape-based Method

With the continuous development and improvement of the demand for deep-space exploration missions and propulsion technology,continuous low-thrust transfer technology has attracted more and more attention in the aerospace field.It has the advantages of high impulse,small size,high payload ratio and multiple ignition.Thus,it has been paid more and more attention in the increasingly frequent deep-space exploration missions.It is also one of the important development directions of orbit transfer technology for deep-space exploration missions in the future.And in recent years,many international deep-space exploration missions have adopted continuous low-thrust technology as the main means of orbit transfer.However,the transfer trajectory under low-thrust is highly non-Keplerian orbit,and the traditional conic splicing-curve method cannot be used,which brings problems and challenges to the design and optimization of the transfer trajectory under low-thrust.In this thesis,the design and optimization of low-thrust trajectory based on shaped-method is studied.Specific research contents are as follows:Firstly,the low-thrust transfer trajectory is nonlinear and cannot be solved analytically,and the traditional numerical method is difficult to determine the initial value to make it converge.In this thesis,the initial design of the low-thrust transfer trajectory is carried out based on the inverse polynomial method.The transfer trajectory of the spacecraft can be approximated by the curve fitting of an inverse polynomial with similar shape.On the premise that the number of orbits of the spacecraft is determined,there is an analytical relationship between the shape parameters of the inverse polynomial and the launch time and transfer time of the spacecraft.At this time,the core of the initial orbit design of the low-thrust spacecraft is transformed into the search for the optimal launch time and transfer time.In order to solve the problem and ensure the rapidity of parameter optimization,the optimization process and parameter selection of genetic algorithm are discussed in this thesis,and the method is successfully applied to the solution of initial orbit design.However,in order to ensure the convergence speed,the integral solution of inverse polynomial method is approximated by equal-segment trapezoidal integral formula,which makes the accuracy of the solution result not ideal.Then,aiming at the inaccuracy of inverse polynomial method mentioned above,this thesis designs a hybrid algorithm that combines BP neural network algorithm and genetic algorithm.Firstly,1000 groups of sample points were calculated by using the fourthorder Runge-Kutta integral solution formula with higher accuracy provided by MATLAB,which were sent to BP neural network for training.Then,the predicted value of neural network was used as the fitness of genetic algorithm for parameter optimization.For the part of BP neural network algorithm,the model and weight updating process of BP neural network algorithm as well as the determination of relevant parameters of algorithm are discussed.The simulation results show that,compared with the classical genetic algorithm,the hybrid algorithm can successfully improve the accuracy of the inverse polynomial method when solving the initial orbit design parameters.Finally,the Gauss pseudospectral method is used to further optimize the transfer orbit of the low-thrust spacecraft.In the process of optimization,the thrust was taken as the control variable,and the optimal fuel was selected as the performance index.The initial design results based on the classical genetic algorithm and the hybrid algorithm were put into the Gauss pseudospectral method respectively,and the transfer orbit of the spacecraft,the change curve of thrust and the change curve of velocity during the transfer process were successfully solved.It is further proved that the initial design parameters obtained by the hybrid algorithm are more accurate than those obtained by the classical genetic algorithm.