Trajectory Optimization Method for Flexible Deployable System based on Conditional Recurrent Neural Network Surrogate Model

Large deformation and rotation are nonlinearly coupled in the motion of flexible deployable system. The deployment trajectory of flexible deployable system is hard to be solved efficiently, which increases the difficulty of trajectory optimization. In this paper, a dynamic trajectory optimization method for the flexible deployable system is proposed based on a conditional recurrent neural network. The mathematical description of the dynamic trajectory optimization problem is derived for a typical flexible deployable system. The dynamic surrogate model of flexible deployable system based on the conditional recurrent neural network is established to circumvent the time-consuming numerical integration in the trajectory optimization. For solving the discontinuity problem of the surrogate model, the basic trajectory sequences are introduced and combined with non-gradient optimization algorithms. The dynamic trajectory optimization algorithm is developed for flexible deployable system with a nonlinear objective function. A flexible deployable array system is taken as the case, and the maximum strain energy of thin plates in the deployment process is the objective function. The numerical results and experimental data demonstrate that there is 26.33% reduction in the objective function for solving the trajectory optimization problem of flexible deployable system with the nonlinear optimization target.