考虑重力影响的柔性关节空间机械臂任务空间神经网络控制

针对柔性关节空间机器人地面模拟及空间应用受到重力或微重力不确定项干扰的问题，提出一种基于奇异摄动的任务空间神经网络自适应控制算法。首先建立了考虑重力影响的柔性关节空间机器人的模型，通过奇异摄动方法将系统降阶处理后，转换到任务空间进行控制器的设计，利用神经网络自适应算法对系统的不确定性进行逼近，并针对逼近误差设计了鲁棒控制器补偿，最后将控制量转换为关节空间的控制力矩。对系统进行了稳定性分析和仿真校验，结果表明，控制器可以解决柔性带来的系统高频谐振和失稳问题，并能有效抵抗系统扰动，具有很好的跟踪性能。

Adaptive Network Control of Flexible Joint Space Manipulator in Task Space under Gravity Effect

Aiming at trajectory tracking of flexible-joint space robot affected by gravity uncertainties, a neural network adaptive control algorithm in task space is proposed for flexible-joint space robot based on singular perturbation theory in this paper. Fristly the model of flexible-joint robot is established, then the system order is reduced by using the singular perturbation method and the dynamics is transformed from joint space to task space, and an adaptive neural control algorithm is designed to approximate the system uncertainty, last the control quantities as the motor control torques is transformed from task space to joint space. The stability analysis is presented, and the effectiveness of the scheme is verified by the corresponding simulation results. The controller can solve both high frequency resonance and instability problems caused by flexible-joint, and also it can resist the disturbance such as gravity effectively and has good tracking performance .