This paper explores the design of an anti-saturation adaptive finite-time control strategy with the neural network (NN) technique for the space circumnavigation mission. Before executing the controller design, the analytical solutions of the desired angular velocity and its derivative of the active spacecraft are calculated. Since there are uncertain saturation constraints on control forces and moments in the actual propulsion system, an auxiliary system compensated by an adaptive NN is adopted. The modified auxiliary system no longer needs the precise output values of the actuators. Besides, the hyperbolic tangent function is introduced to design the weight update law for the NN compensator, so that the derivative of the weight estimator will not be amplified by the quadratic of states when the system states are large. It is proved that tracking errors of the system states can converge to a residual set of the origin in finite time. Simulation results show that the maximum amplitudes of the control signals are greatly reduced compared to the classical non-singular terminal sliding-mode control scheme, and that the neural-based compensator can significantly weaken the overshoot and chattering.

本文探索了用于神经网络绕行飞行任务的抗饱和自适应有限时间控制策略的设计。在执行控制器设计之前，需要计算活动航天器的期望角速度及其导数的解析解。由于实际推进系统中控制力和力矩存在不确定的饱和约束，因此采用了由自适应神经网络补偿的辅助系统。修改后的辅助系统不再需要执行器的精确输出值。另外，引入双曲正切函数设计神经网络补偿器的权重更新定律，使得系统状态较大时，权重估计量的导数不会被状态的二次态放大。证明了系统状态的跟踪误差可以在有限时间内收敛到原点的残差集。仿真结果表明，与传统的非奇异终端滑模控制方案相比，控制信号的最大幅度大大减小，并且基于神经的补偿器可以显着减弱过冲和颤动。