机械臂的CONN控制与CMAC控制的比较研究

由于正交神经网络算法简单，学习收敛速度快，具有线性、非线性逼近精度高等优异特性，取得了较好的应用效果，但在机器人动态建模与实时控制问题上研究较少。为此在机械臂的神经网络控制中，该文提出复合正交神经网络(CONN)与PID并行控制方法，并对小脑模型(CMAC)与PID并行控制作一比较研究。仿真结果表明，当阶跃输入与正弦输入时CONN与CMAC实现的前馈控制具有相同的控制效果，但CONN算法比CMAC算法更简单，这充分地体现了复合正交神经网络的特点。

Comparison between CONN Control and CMAC Control of the Manipulator

The orthogonal neural network obtained better application results because of its simple algorithm,high-speed convergence of learning process, and excellent characteristics in the linear and nonlinear accurate approximation. But few of research works have been done for dynamic system modeling and real-time control in robot neural network controls. Therefore, a parallel control method of compound orthogonal neural network(CONN) and PID is presented, and is compared with a parallel control of CMAC and PID in manipulator neural network controls. The simulation results prove that both the former and the latter have the same control effects when signals of step wave and sine wave are inputted in the feedforward control. But CONN algorithm is simpler than CMAC algorithm.